JP2002197410A - Information reproducing system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To correctly reproduce information by reading a densely recorded code pattern. SOLUTION: From among a plurality of blocks, a block to which a block receiving the small influence of optical distortion is selected and read. When a code is recorded by double heights, for example, data in a processing object block 82 is read in the order 1 to 9. Since the block of 1 is closer to the end of a frame 60 even though the same data is recorded in the block of 1 and the block of 2, the probability of the occurrence of a read errors due to optical distortion is high. Then, by reading the data in the block of 2 once more and overwriting it after that, more precise data can be obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to so-called multimedia including audio information such as voice and music, video information obtained from cameras and video equipment, and digital code data obtained from personal computers and word processors. The present invention relates to an information reproducing system that optically reads the above code pattern from an information recording medium such as paper on which information is recorded as an optically readable two-dimensional code pattern and reproduces the original multimedia information.

[0002]

2. Description of the Related Art Conventionally, various media such as magnetic tapes and optical disks have been known as media for recording voice, music and the like. However, even if these media are made in large quantities, the unit price is somewhat expensive, and a large space is required for storage. Furthermore, if it is necessary to hand over the recorded audio medium to another person at a remote location, it takes time and effort to mail it or take it directly. There was also. Also, other than audio information, video information obtained from cameras, video equipment, etc., and digital code data obtained from personal computers, word processors, etc.
The same applies to the so-called multimedia information as a whole including the above.

To cope with such a problem, Japanese Patent Laid-Open Publication No. Hei 6-231466 discloses that multimedia information including at least one of audio information, video information, and digital code data can be transmitted by facsimile. In addition, a plurality of dots are two-dimensionally arranged as image information that can be copied in large quantities at low cost, that is, encoded information.
A system for recording on an information recording medium such as paper in the form of a dimensional code and a system for reproducing the same are disclosed.

In the information reproducing system disclosed in this publication, an information reproducing apparatus for optically reading and reproducing a two-dimensional code on an information recording medium is held by hand, and is recorded on the recording medium in accordance with the recorded two-dimensional code. A method of reading by manually scanning is disclosed.

[0005]

The structure of the two-dimensional code pattern itself, which can further improve the recording density, has been studied, and it is expected that higher density recording will be achieved.

[0006] However, the above publication does not yet give sufficient consideration to reading such future high-density recorded code patterns.

An object of the present invention is to provide an information reproducing system capable of reading a code pattern recorded at high density and reproducing information correctly.

[0008]

In order to achieve the above object, an information reproducing system according to the present invention provides an information reproducing system comprising a plurality of blocks each including a data code corresponding to data of information to be reproduced. An information reproducing system for optically reading the code and reproducing the information from an information recording medium recorded so as to be readable, wherein the image input means captures the code and the image input means captures the code. Data reading means for reading the data from the image, and reproducing means for reproducing the information from the data read by the data reading means, the data reading means in the image taken by the image input means A block configured to select and read a block having a small influence of optical distortion from a plurality of the above blocks. Characterized in that was.

That is, an image obtained by capturing a code is distorted due to optical distortion. The effect of this distortion increases as the code density increases, and the possibility of a data code reading error increases. I do. Therefore, according to the information reproducing system of the present invention, a block having a small influence of optical distortion is selected and read from a plurality of blocks in a captured image. Therefore, the probability of occurrence of a reading error can be reduced.

[0010]

Embodiments of the present invention will be described below with reference to the drawings.

FIGS. 1A to 1C show examples of two-dimensional codes (dot codes) recorded optically readable on an information recording medium.

That is, as shown in FIG. 1A, the dot code is composed of a data code 10 and a pattern code 12. The data code 10 indicates contents corresponding to actual multimedia information. The pattern code 12 is for determining a reference point for reading the data code 10 and includes a pattern dot 14 and a marker 16. Here, the pattern dot 14 is an individual dot of one or a plurality of dot groups other than the marker constituting the pattern code 12.

The dots constituting the data code 10 (hereinafter referred to as data dots) are recorded at a very high density (for example, near the resolution limit of an image pickup device such as a CCD). In order to correctly reproduce, each data dot must be read accurately, that is, the reading position must be accurately detected. Therefore, in order to accurately know the relative positional relationship between the pixels of the CCD and the data dots thereof, they are arranged at predetermined positions with respect to the data code 10 and have predetermined dot patterns (black is 1, black is 0, white is 0). Then, 1010
The pattern dots 14 having the above 1), which are slightly larger than the data dots, are used. However, since it is difficult to find the pattern dot 14 from nothing, a very large marker 16 is provided at a predetermined position with respect to the pattern dot 14. In this example, the marker 16 is circular, and its diameter is (marker 16)> (pattern dot 14)>
(Data dot).

Of course, the pattern dots 14 may be recorded in the same size and shape as the data dots as shown in FIG. Also, the marker 16 may be recorded in a special shape, color, or size different from the pattern dots 14 and the data dots, for example, as shown by the arrow shape shown in FIG.

As shown in FIG. 1C, a data code 10 is arranged in a rectangular area, a pattern code 12 is arranged above and below the data code 10, and address dots 1 are arranged on both sides.
8 may be arranged to form a block, and such blocks may be arranged in a matrix to constitute a two-dimensional code. Here, for example, the pattern dot 14 can be expressed as “1001001” if the black dot is represented by 1 and the white dot is represented by 0.
[00100100001001001001001]. Also, address dot 1
8 is, for example, “1 ****************
#### 0 ". In addition,
“*” Represents 1 or 0, and an address is represented by these 10 addresses. “#” Represents 1 or 0, and 10
The error correction code of the address is indicated by the number.

As described above, in the information recording medium on which the dot code including the data code 10 corresponding to the content of the multimedia information to be reproduced and the pattern code 12 for determining the reading reference point is recorded,
The pattern code 12 is replaced with a pattern dot 14 arranged at a predetermined position with respect to the data code 10.
And a marker 16 which is arranged at a predetermined position with respect to the pattern dot 14 and is used for detecting the pattern dot 14, so that the pattern dot 14 forms the data code 10. The same data dots can be detected based on the marker 16, and the read reference point can be accurately obtained from the detected pattern dots 14.

In the case where the marker 16 is divided into blocks as shown in FIG. 1C, the markers 16 do not have to be arranged at the four corners of the block. For example, as shown in FIGS. As shown, a marker 16 can be placed.

That is, FIG. 2A shows a rectangular block 2
The marker 16 is arranged on each side of 0.

FIG. 2B shows an example in which, when the blocks 20 are arranged in a plurality of rows, the markers 16 are arranged only on the top row.

FIG. 2C shows an example in which, when the blocks 20 are arranged in a plurality of stages, the markers 16 are recorded with some parts omitted.

FIG. 2D shows a case where the marker 16 is recorded in a square shape. As described above, even when the shape of the marker 16 is not circular, various marker arrangements can be adopted as shown in FIGS.

Preferably, the marker 16 is circular for the following reasons. That is, if the shape of the marker 16 is circular, there is no deformation due to the rotation (skew) of the dot code and the image pickup device (camera) such as a CCD, so that even if the skew occurs, the marker 16 can be easily extracted. Because it can be. However, when the skew is within a certain range, for example, when imaging is performed in a state where the relative relationship between the camera and the dot code is defined, the marker 16 may be other than a square or a hexagon. May be used. Conversely, if such an anisotropic shape is used, if one marker 16 is detected,
There is also an advantage that the skew angle, that is, the direction in which the skew is inclined, can be obtained only by looking at the shape.

FIG. 2E shows an example in which the block 20 is formed in a hexagonal shape and the markers 16 are arranged at the three vertices. Thus, the shape of the block is not limited to a rectangle.

As described above, a plurality of blocks in which the data code 10 composed of a plurality of dots is arranged in a predetermined area are arranged two-dimensionally to form a dot code, and the marker 16 is set to the relative position of each block. As you can see the relationship,
Since they are arranged in a part of each block, these markers 1
6 facilitates extraction of block boundaries.

Further, since the approximate center of the marker 16 can be obtained by extrapolation from the marker pair as will be described in detail later, the marker 16 is not required at all four corners of the block.
Therefore, as shown in FIGS. 2B and 2C, some of them can be omitted. In this case, this omitted location
There should be only dots or nothing at all.

FIGS. 3A and 3B are examples in which the point 22 is recorded at the omitted position. In this example, vertices forming opposite sides that are different between blocks adjacent in the first direction, and
Markers 16 are arranged at vertices forming the same opposite side between blocks adjacent in the second direction. In short, the upper marker is recorded in the direction adjacent to the horizontal direction in the figure,
On the other hand, when the marker is adjacent in the downward direction, the lower marker is recorded. When the marker is adjacent to the lower marker, the upper marker is recorded.
, So that the markers 16 are striped.
Is recorded (or omitted).

As described above, the vertices forming the opposite sides between the rectangular blocks adjacent in the first direction and the vertices forming the same opposite side in the blocks adjacent in the second direction orthogonal to the first direction are marked with the marker. By arranging the markers 16, that is, by arranging the markers 16 in a stripe pattern, the markers 16 that require a larger area than the case where the markers 16 are arranged at the four corners of the block 20
And the data dots can be arranged at the omitted marker positions, and the recording capacity can be increased accordingly.

FIG. 4 shows an example in which the markers 16 are arranged in a checkered pattern (or omitted).

By arranging the markers 16 at vertices at different diagonals between adjacent blocks (that is, in a checkered pattern), the markers 16 are arranged at four corners of a rectangular block as compared with the case where markers are arranged at four corners of a rectangular block. The number of markers requiring a large area is reduced to half, and data dots can be arranged at the omitted marker positions, so that the recording capacity can be increased accordingly. In addition, since at least one set of markers 16 can be always detected on one line, there is also an effect that the markers 16 are more resistant to distortion than when the markers are arranged in a stripe pattern as described above.

Next, the pattern dots 14 will be described.

As shown in FIG. 5A, for example, as shown in FIG. 5A, if a black dot is represented by 1 and a white dot is represented by 0, "10101011010010101" is obtained.
For example, it is necessary to arrange adjacent black dots such as “0101” so that they can be separated from each other. That is, the black dots are arranged at a distance. Of course, the black dots and the white dots need not be recorded alternately, as long as there is a space between the black dots and the black dots. Similarly, a space is provided between the data dots and the pattern dots so that the black dots are spaced apart.

This is because if two black dots constituting the pattern dot 14 are adjacent to each other, for example, two black dots are set to one in the determination process of the reading reference point by the pattern dot 14 as described later in detail. Is detected as if it were a black dot, which could be a source of error. That is, in the process of determining the reading reference point,
For the pattern dot 14, the center coordinates of each black dot are obtained for each dot, and the read reference point is determined by using the center coordinates. Therefore, for the pattern dot 14, the center point position of each dot needs to be obtained. , Black dot 2
If they are adjacent to each other, the correct dot center position cannot be obtained.

That is, as described above, by arranging the pattern dots in such a manner that adjacent dots are separable from each other, the pattern dots can be formed by relatively simple calculations such as the center of gravity and the center of a circumscribed rectangle. The center of each dot to be obtained can be obtained accurately. Further, even when information (address information, header information) is recorded on the pattern code, the black dots are not linked and cannot be separated.

Further, by performing some kind of modulation so that a sequence such as “1” or “1” does not appear,
Even if the dot recording positions are arranged adjacent to each other, it is also possible to superimpose and record some data such as address information and header information in the pattern dots 14.

For example, 4-bit / 6-bit modulation as shown in FIG. In this case,
The input signal 0 (that is, 0000) is 0 (that is, 0000).
00) instead of 36 (ie, 10010
0). This means that at least one black dot
This is because if there is no individual, it does not make sense as a pattern dot.

As described above, by recording the dot code as a code modulated so that adjacent dots can be separated from each other, even if information (address information, header information) is recorded on the pattern code, the black dot is recorded. Can be prevented from being connected and unable to be separated.

The format of the pattern dot 14 may be various other forms, but it is necessary that the pattern dot 14 be a special code having a positional relationship or a structure capable of estimating the position of the reading reference point. With such a special code, the reading reference point can be easily estimated using the special code.

For example, as shown in FIG. 6B, a linear pattern dot 14, that is, a pattern line may be used. As a straight line detection algorithm, a method of obtaining a straight line by expressing a straight line by parameters of θ and ρ and performing counting on a θ-ρ plane as is well known in image processing (a straight line detection algorithm, for example, Hough transform) is known. The intersection of a straight line extracted using this method can be used as a reading reference point.

Alternatively, as shown in FIG. 5B, the pattern dots 14 may be formed by providing at least a part of the recording dots of the pattern dots with a phase shift. By providing such a phase shift, the pattern dots 14
Becomes stricter, so that the accuracy of estimating the reading reference point can be improved.

That is, originally, the recording positions of the pattern dots 14 and the dots of the data code 10 are arranged substantially on a lattice, but in the case of the pattern dots 14, they are slightly shifted from the lattice points. This is called a phase shift.

For example, in FIG. 7 (A), (1) shows a case of an equi-pitch arrangement, which can be detected by reading at any point within a dot. If a dot is read at this equi-pitch point, it cannot be detected unless there is a read point in the hatched portion in the figure. Therefore, the range of the read reference points from which all the dots can be detected is narrower than in the case where no phase shift is provided, and the estimation accuracy of the read reference points can be improved.

Further, by providing such a phase shift, it is possible to estimate the shift of the reading reference point position from the positions of undetected dots and their deviations.

For example, when a phase shift is provided, if a certain dot cannot be detected but another dot can be detected, the type of the phase shift of the undetectable dot is determined. Since it is possible to know from the preset format information, it is possible to estimate the direction in which the position of the reading reference point is shifted based on the phase shift.

As shown in FIG. 7B, a pattern dot 14 for detecting a reference point for reading the data code 10 is provided between the markers 16 as the pattern code 12.
A guide code 26 which is smaller in size than the normal dots of the pattern dots 14 and which has a phase shift of up, down, left and right added to the dot position reference point 24 of the pattern dots 14
May be provided.

In the case of such a pattern code 12, after the pattern dot 14 is read, the marker 16
The center position is moved in the vicinity so that the guide code 26 can be read by connecting points shifted one dot up and down from the center of. By doing so, the reading condition of the guide code 26 can be reduced without increasing the reading condition of the pattern dot 14 and the reading reference point (in this case,
The accuracy of estimation of the marker center position) can be improved.

As described above, by recording at least a part of the pattern dots with a small diameter, many pattern dots can be recorded in the same space, and the recording density of the pattern dots can be increased. In addition, as the detection conditions for pattern dots become more strict, the accuracy of estimating the reading reference point can be increased.

As shown in FIGS. 5B and 5C, a part of the dots 14 from the center of the pattern dots 14 is formed.
The diameter of B may be recorded larger.

That is, when the pattern dots 14 are detected on a line connecting the approximate centers of the marker pairs as described later, the dots 14B from the center where the detection is most difficult are preliminarily recorded so that the pattern dots 14 can be detected. It is possible to eliminate losing, and the dot 14B from the center
Can be detected first, and another dot can be roughly searched.

Next, the address dots 18 will be described.

The address dots 18 are provided between the markers 16 in a direction orthogonal to the pattern dots 14 as shown in FIGS. 8A and 8B, for example, and indicate the address of the block 20.

By providing such address dots 18, even when blocks are not read in order,
By rearranging the read blocks in the order of the addresses indicated by the address dots 18, the blocks can be correctly reproduced.

Further, as shown in FIG. 8C, if an inverted code 28 obtained by inverting the address dot 18 is recorded on both sides of the address dot 18,
Using this, error correction and correction of the reading position can be performed.

Alternatively, as shown in FIG. 8D, by providing a correction code 30 for error correction on both sides of the address dot 18, it is also possible to perform address error correction. That is, in addition to the address dot 18,
By reading this correction code 30, an address error can be corrected.

A part of the address dot 18, for example, as shown in FIG. 9A, a fixed pattern 32 which is always a black dot, that is, "1", in the middle regardless of the address. It may be provided. In this way, by reading the black dots of the fixed pattern 32, it can be used for detecting a positional shift or detecting an error.

Alternatively, the fixed patterns 32 may be provided at both ends as shown in FIG. In this case, as the fixed pattern 32, by setting one end of the address dot 18 as a black dot and the other end as a white dot, the reading direction of the address dot 18 can be specified. Even if is read upside down, the address can be read accurately.

As described above, if the fixed pattern 32 is included in at least a part of the address dot 18, alignment and error detection can be performed using the fixed pattern 32. The reading direction can be specified.

When the blocks 20 are recorded in the order of the addresses, the addresses of a certain block can be easily estimated from the addresses of the surrounding blocks, so that the address dots 18 can be omitted as appropriate. is there.

For example, as shown in FIG. 9C, when the blocks 20 are aligned and sequentially assigned addresses, if the addresses of the first block and the third block can be detected, the addresses of the blocks between them are detected. Can be easily estimated to be the second block. Therefore, the address of the second block does not have to be recorded, and the data code 10 can be recorded there, so that the recording capacity can be increased accordingly.

If the address relationship between adjacent blocks is set in advance, for example, as “+10”,
The block addresses need not be arranged in order as shown in FIG.

As described above, if the address dots 18 are appropriately omitted and recorded so that the addresses of at least two or more blocks can be determined simultaneously, the data code 10 can be recorded even in the omitted portion, and accordingly, Thus, the recording capacity can be increased.

Next, an information reproducing system for optically reading the dot code described above and reproducing the original multimedia information will be described.

FIG. 10 is a diagram showing the configuration.

That is, the dot code 36 recorded on the information recording medium 34 such as paper is picked up by the image input unit 38. The data reading reference point determination unit 40 determines a reading reference point for reading the data code 10 from the image captured by the image input unit 38. The data reading unit 42 is configured to output the image input unit 3 based on the reading reference point determined by the data reading reference point determining unit 40.
The data code 10 is read from the image obtained in step 8. The data reproducing unit 44 reproduces the data read by the data reading unit 42.

With such a configuration, the dot code can be read at high speed and accurately.

The reproducing unit 44 reproduces the read data by performing demodulation and error correction as necessary. Here, "reproduction" refers to video output to a monitor for an image, and audio output to a speaker for audio. If the signal is a compressed signal, it is decoded and then output. For example, if a Reed-Solomon triple product code is used for error correction, since data is 8-10 modulated, 10-8 modulation (8-10 inverse modulation) is performed.

On the other hand, the image input section 38 includes an imaging section 46 such as a CCD camera, and an equalization / binarization section 48 for equalizing and binarizing an output image of the imaging section 46. That is,
In the image input unit 38, the image obtained by imaging the dot code 36 is binarized, and the subsequent data reading reference point determination unit 40 processes only the binarized data. As a result, not-shown memory for holding images can be reduced, and the amount of data transfer is also reduced, enabling high-speed processing.

If such binarized data is used, the data reading reference point determining unit 40 and subsequent units can be constituted by an information processing device such as a personal computer. Further, in such a case, the size of the housing that accommodates only the image input unit 38 may be small, so that the burden on the operator when scanning is reduced.

On the other hand, the data reading reference point determining unit 4
For example, 0 may detect the middle point of the pattern dot 14 as a reading reference point, but in the present embodiment, the center position of the marker 16 is determined as the reading reference point.

That is, the data reading reference point determining unit 4
0 denotes a marker extracting unit 50 for detecting a marker existing area where the marker 16 is present from the image data binarized by the equalizing / binarizing unit 48 and an outline of the marker 16 in the detected marker existing area. A marker detecting section 54 comprising a marker approximate center calculating section 52 for obtaining the center;
The pattern dot 14 is detected by using the approximate center of the marker 16 obtained in step 4, and the detected pattern dot 1
4 and a marker center detector 56 for estimating the true center coordinates of the marker 16 as a reference point for reading the data code 10 and the address dot 18.

The marker extracting section 50 of the marker detecting section 54
Performs erosion and labeling,
Detect the marker area.

Here, the erosion process itself is a process known in image processing, and there are various types of hardware for executing the process. This erosion processing is shown in FIG. 11B when a target pixel is masked with a mask of a predetermined size, for example, a mask of 3 × 3 pixels as shown in FIG. If all the pixels in the mask are black pixels (indicated by ○), the pixel of interest is a black pixel, and as shown in FIGS. 11C and 11D, even one white pixel (× If this is the case, the pixel of interest is converted into a white pixel.

When such erosion processing is performed on the dot code 36, as shown in FIG. 12, the data code 10 composed of small dots and the pattern dot 1
4. The address dots 18 have disappeared, and only pixels corresponding to the markers 16 which are much larger than these dots can be left.

Then, the remaining pixels are sequentially labeled for each group, and an existing area having the same label is obtained. The area reduced by erosion into this existing area is enlarged to become a marker existing area 58.

The marker approximate center calculation unit 52 calculates the center of gravity of the black pixel for each marker present area 58, and uses this as the reference point for reading the pattern dots 14 with the marker approximate center.

As described above, by performing the erosion processing and the labeling processing, the marker 16 can be easily detected by using existing hardware, and the influence of noise near the marker 16 can be reduced. .

Further, the marker extracting section 50 can obtain the marker area 58 without performing the above-described processing requiring calculation of streak and the like known in erosion and image processing. .

For example, considering a frame image as shown in FIG. 13A, an image is taken in a state where an LPF is applied due to the characteristics of an optical system (not shown) of the imaging unit 46.
As shown in (B), a black level is detected to be low in a portion where black pixels gather like the marker 16. Therefore, a minimum value of luminance is obtained while scanning an area including at least one or more markers of the frame 60, and a value obtained by multiplying the minimum value by an offset (about 15 to 20) is set as a threshold th.
When the frame image is binarized, the marker central portion 62 can be easily extracted as shown in FIG.

That is, since the marker 16 is large, the luminance value of the central portion takes a considerably low value when the marker 16 is captured as an image. On the other hand, the dots of the data code 10 and the pattern dots 14 are slightly blurred due to slight blur due to the optical system, and the surrounding white information is superimposed thereon. Therefore, since a luminance profile as shown in FIG. 13B is obtained, only the center portion of the marker 16 can be extracted by binarizing it with the predetermined threshold th.

As described above, the minimum value of the luminance is obtained for at least a part of the frame 60, and binarization is performed using a threshold value obtained by adding a predetermined offset to the minimum value, so that calculations such as erosion and streak are not required. The center part 62 of the marker can be detected.

In this case, if this processing is performed after applying a low-pass filter, the marker central portion 62 can be detected more easily. That is, when a low-pass filter is applied to a frame image as shown in FIG.
Small dots such as data dots are further blurred and thin, and only the center of the large marker 16 remains.

As described above, by applying the low-pass filter, when the marker 16 is recorded as a dot larger than other dots such as the data dot of the data code 10 and the pattern dot 14, the detection of the marker central portion 62 is performed. Can be performed more reliably.

Alternatively, the center calculation range may be obtained from the number of streaks that are in contact with the streak processing, which is a technique known in the field of image processing.

That is, as shown in FIG. 14A, a streak (determining the number of continuous black runs, that is, the number of continuous black pixels and its starting point) is performed for each frame, and the streaks (see FIG. In the example, the streaks S1 to S7) are put together. Then, the combined streak copies the attribute data. Here, the attribute data refers to the coordinates of the upper end, the left and right ends of the streak and the streaks in contact therewith, and the number of streaks in contact. And when the streak that comes in contact no longer exists below,
The range in the height direction is obtained from the number of streaks that have come in contact with it.

In this manner, when the data is decomposed into the data of the starting point of the black pixel and the length from the starting point, small dots such as data dots take only short streak data.
Only the streaks that are in contact with a lump of a certain size can be determined to be the marker 16.

Therefore, the streaks that are in contact with each other are grouped together, and the attributes of the grouped streaks, that is, the coordinates of the start point and the length (end point) are copied to the lower streaks. If there is a streak in contact, the lower streak copies the upper attribute, so the start and end points of the upper attribute are copied. Thus, the left end, the right end, the upper end, and the lower end are copied.

For example, the uppermost small streak S1
Performs the copying of only the uppermost data, that is, only the upper side, to the lower streak S2,
2 has left end and right end data. Then, when the processing is finally completed up to the lowest streak S7, the attributes of the lowest streak data S7 include the left end and the right end of the longest streak S4 and the top streak S1. And the attribute of the bottom streak S7 are recorded, and the area where the marker 16 is present is obtained from this attribute.

As described above, if the center calculation range is determined by the number of streaks that are in contact with the streak processing, processing that requires a large amount of calculation, unlike the erosion processing, does not need to be performed, thereby simplifying the processing. The marker 16
It also has the effect of increasing the resistance to spot-like scratches inside.

Further, it is possible to extract all the markers without performing the above-described marker extraction processing on all of the frames.

That is, as shown in FIG. 14B, when there are nine markers 16 (M1 to M9) in the frame 60, the marker 16 is first raster-scanned in the frame. Then, the marker 16 is searched for by any of the marker extraction processes described above. When the marker M1 is found in this way, the markers M2 and M4 relatively adjacent to the marker M1 are given the information indicating how far away from the marker M1 the information is given as the look-ahead information. Further, it is not necessary to perform a raster scan to search for the next markers M2 and M4 in detail, and it is necessary to take into account the possibility of skew and jump to an area 64 where a marker determined in accordance with the look-ahead information may exist. Only in the area 64, the marker 16 need be searched. And if the markers M2 and M4 are found,
From there, the marker 16 may be searched for by expanding the marker one after another.

As described above, at least one or more markers are detected, and a marker is searched only in the area 64 where there is a possibility that another marker may exist based on the detected marker. The calculation for marker extraction is not required for the whole, and the processing speed is increased. If there is black dust between the markers 16, for example,
In the process such as the erosion described above, such a dust is also extracted as a candidate for the marker 16, but when searching only in the area 64 where the marker may be present, such dust is extracted as a marker candidate. Never.

In the marker extraction process, for example, as shown in FIG.
When only a single marker is found, it is known that a marker set (four corners in this example) constituting the block cannot be obtained, so that the processing of this frame may be stopped.

That is, in the dot code in which the markers 16 are arranged at the four corners, if up to three labeled labels have been touched after the erosion process or the streak process, there are only three markers 16. It turns out at that point. Therefore, when the number of markers detected in a frame is less than the number necessary to form a block, the processing of that frame is stopped, so that useless processing can be omitted.

On the other hand, the marker approximate center calculating section 52 calculates the center of gravity of all black pixels in the marker existing area 58 as shown in FIG. The approximate center of the marker is used as the reference coordinate for the detection of.

As described above, by setting the center of gravity of all the black pixels in the marker existing area 58 as the approximate center of the marker, even if there is a white spot, a black dot, or a whisker-like noise, it can be canceled.

Alternatively, the marker approximate center calculating section 52
Is, for example, as shown in FIG.
The circumscribed rectangle 68 may be obtained, the center point 70 thereof may be calculated, and the center point 70 may be set as the approximate center of the marker. In this case, the amount of calculation can be reduced as compared with the case where the center of gravity is calculated.

In the marker approximate center calculating section 52, for example, as shown in FIG. 16 (A), when the marker 16 is Approximate center 74 of the remaining portion indicated by hatching in FIG. If the calculated approximate center 74 deviates from the actual correct marker approximate center 76, then the marker center detection unit 5
In the detection of the pattern dot 14 performed by Step 6,
On the basis of this point, the pattern dots 14 may not be detected properly. In other words, when the marker 16 is over the frame end 72, a part of the marker 16 is lost, so that the detected approximate center position of the marker may be shifted. In such a case, each dot of the pattern dot 14 is normally replaced. It cannot be detected, and the marker center position detection accuracy is reduced.

The marker approximate center calculating section 52 estimates whether or not the marker 16 is missing based on how far the calculated approximate center position is from the frame end 72. In such a case, it is preferable to perform a process of calculating the shift amount δ and correcting the approximate marker center position to the correct approximate marker center 76.

In this case, whether or not the marker 16 is missing can be determined based on whether or not the calculated approximate center position is within the approximate center correction area 78, as shown in FIG.

In the approximate center correction area 78, the distance d from the frame end 72 of the correct marker approximate center 76 to the right and left ends of the frame is smaller than half a of the size of the marker 16 in the left and right direction (that is, d < a) With respect to the upper and lower ends of the frame, the distance d from the frame end 72 of the correct marker approximate center 76 is smaller than half the vertical size b of the marker 16 (ie, d <b). It is an area with a different height.

The deviation amount δ of the approximate center is given by the following equation.

[0101]

(Equation 1)

Since the pixels of the CCD of the image pickup section 46 are not square, and the pixel pitch of the CCD is different between the vertical and horizontal directions, as shown in FIGS. 16A and 16B, The image obtained by imaging the circular marker 16 is thus elliptical. Therefore, the formula for calculating the shift amount δ differs between the case where the marker 16 protrudes from the left and right ends of the frame and the case where the marker 16 protrudes from the upper and lower ends.

As described above, by correcting the approximate center position of the marker that may be missing at the end of the frame, the approximate center position of the marker at the end of the frame assuming that the marker is not missing is also corrected. Since it can be obtained, the detection of the pattern dot can be made more reliable.

Even if the above-mentioned deviation amount δ, that is, the correction amount, is not calculated sequentially, it is necessary to calculate this formula in the form of a table showing the relationship between the detected distance from the approximate center frame end and the correction amount. Good to have.

For example, the marker approximate center calculating section 52 performs the processing shown in FIG.
As shown in FIG. 6 (C), there is a table showing the relationship between the detected distance from the approximate center frame end and the corrected distance from the approximate center frame end. However, in these tables, the diameter of the marker 16 is 7 dots, the size of one dot at the time of imaging is 3.47 pixels (pixel) in width, and 2.70 in height.
The case of a pixel is shown.

As described above, if a table for outputting a predetermined correction amount according to the distance from the frame end is used, the calculation of the correction amount becomes unnecessary.

It is not necessary to perform the above-described approximate center position correction processing on the markers of the entire frame, but only on the markers related to the block 20 for actually reading the data dots.

For example, in FIG. 17A, in the frame 60, the marker approximate centers M0 to M0 indicated by “x” marks are respectively provided.
When M14 is calculated, there are three markers M0, M9, and M14 that need to be corrected for the approximate center position, that is, the markers 80 that are partially missing at frame ends. However, the block to be actually read, that is, the processing target block 8
Reference numeral 2 denotes five blocks shaded in the figure with markers at four corners (markers M7, M9, M11, M
Blocks in which 13 are arranged at the four corners are excluded for the reasons described later in detail). Therefore, when it is determined that some of the markers 16 constituting the five blocks involved in the reading are missing at the frame end, correction is performed only on the markers, and no correction is performed on the other markers. . That is, in the example of FIG.
Since only the marker M14 is involved in the processing target block 82 among the markers 80 partially missing at the end of the frame, the above-described approximate center position correction processing is performed only on the marker M14 (in this case, the marker M14 is used). M0, M9
Cannot be one of the four corner markers constituting the block, and is excluded from the processing target).

As described above, the correction is selectively performed only on the marker to be processed, that is, reading is performed instead of detecting and correcting the marker requiring correction of the approximate center position of the entire frame. Since this process is performed only for the marker related to the block, the amount of calculation is reduced.

The markers M7, M9, M13, M11
Are excluded at the four corners because the markers M5, M6,
Blocks in which M10 and M8 are arranged at four corners and markers M7 and M
9, M13, and M11 have the same data content as the block in which the four corners are arranged.
Because the former block is selected, the latter block is excluded.

Next, the marker center detecting section 56 will be described in detail.

The marker center detecting section 56 first selects two markers 16, that is, a marker pair, in order to detect the pattern dots 14.

That is, when the pattern dot 14 is searched using the marker 16, if only one marker 16 is used and the dot code 36 is skewed as shown in FIG. It is not known where the pattern dots 14 are. On the other hand, as shown in FIG. 17B, if two markers 16 are detected as a marker pair 84, the pattern dot 14 is given as format information such that the position where the pattern dot 14 exists is, for example, between the markers. If so, it is sufficient to search between the marker pairs 84. For example, if one pixel 16 is sequentially searched from one marker 16 constituting the marker pair 84 to the other marker 16, the pattern dot 14 is obtained. Will be found.

As described above, by selecting the marker pair, the detection of the pattern dots 14 is ensured. That is, since the pattern dots 14 are recorded with a predetermined positional relationship from the marker 16, even if the skew occurs, the pattern dots 14 can be detected if at least two marker positions are obtained.

Therefore, the marker center detecting section 56 first
As shown in FIG. 18, one reference marker to be processed is selected by a reference marker selection unit 86, and a marker existence region is calculated by a marker existence region calculation unit 88 according to the specification skew angle and the format information. , Marker candidate detection unit 9
At 0, this paired marker existing area is searched to detect all the paired marker candidates. Next, the base marker pair selection unit 9
A marker to be processed is determined by selecting a pair of marker candidates forming a pair in step 2, and the marker to be processed is selected by the marker set selecting unit 94 to determine a block to be processed. Then, the process proceeds to the detection of the pattern dots 14 between the markers related to the processing target block.

The marker presence area calculation section 88 calculates the marker existence area as follows. That is, FIG.
As shown in (C) of FIG. 7, the range in which a marker to be paired with one reference marker 96 selected by the reference marker selection unit 86, that is, the pair marker existing area 98 is defined by a specification skew angle (see FIG. (In the example, 20 °) and the distance between the markers, that is, the format information.
Therefore, as shown in (D) of FIG. 17, even if the marker detection unit 54 detects the approximate center (indicated by X in the figure) of some marker candidates due to noise or the like, the range of the pair It can be seen that only those that fall within are markers to be paired (paired marker 100), and the others are erroneous markers 102.

The specification skew angle indicates how inclined the information reproducing system is to read the dot code 36. For example, if the image input unit 38 is fixed and the dot code 36 is automatically conveyed and scanned, there is almost no skew, and the specification skew angle is a unit of 1 ° or less. Become. On the other hand, when the operator operates the image input unit 3
In the case where the user manually scans the dot code 36 over the dot code 36, there is a possibility that about 20 ° is inclined. Also, for example, if the image input unit 38 has a structure in which the scanning direction is regulated to some extent by rollers or the like, the skew can be reduced. As described above, in the information reproducing system, the skew angle is determined by the specification depending on the structure.

As described above, when a marker pair is selected, a marker existence area is calculated according to the specification skew angle and the format information, and a marker candidate is detected by searching the marker existence area. Since it is not necessary to perform a process of detecting a marker candidate that forms a pair for the entire frame, search efficiency is improved. Further, since the search is performed by narrowing the left and right adjacent marker existence areas, erroneous marker detection due to noise can be reduced. In other words, even if marker candidates are erroneously detected due to noise, they can be removed due to restrictions on the format of the marker pair.

In the marker center detecting section 56,
The marker position may be calculated by extrapolation.

For example, if one marker pair has already been found, for example, if the markers M8 and M10 have already been detected in some form in FIG. The position is already skewed by the positional relationship between the markers M8 and M10 without performing the process of calculating the marker presence area 98 from the marker M10 based on the specification skew angle and the format information and searching for the range. Since the angle is known and the distance between the markers is known, the extrapolation vector (exv) 104 is extended from the marker M10 according to the angle, and as shown in FIG. What is necessary is just to search for the paired marker in a narrow range.

By doing so, only a very narrow range becomes the marker search area 106, so that there is no possibility of erroneously detecting a marker candidate located outside of the area, that is, it is resistant to noise. In particular, the above-mentioned pair marker existence region 9
Even when there are a plurality of marker candidates in 8, it is possible to select a correct marker candidate as the paired marker 100.

Further, by using such an extrapolation vector 104, it becomes possible to read a code in which the marker 16 is omitted as appropriate. For example, in the above example, even if the marker M11 is omitted, the approximate position where the marker M11 should be located can be determined from the extrapolated vector 104, and the pattern dot 14 between the markers M10 and M11 can be read.

As described above, when the position of the marker is calculated by extrapolation, it is less likely that a marker will be missing due to noise or an erroneous marker will be detected. Further, it becomes possible to read a code in which the marker 16 is omitted as appropriate.

The extrapolated vector (exv) 104 can be used for searching for a vertically adjacent marker, which is calculated from the positional relationship between the previously detected horizontal marker pair.
For example, when two markers 16 are detected as indicated by crosses in FIG. 19C, the positions of the vertically adjacent markers are calculated from the increments Δx and Δy in the x direction and the y direction by the following equation. Is extrapolated.

Exv. fx = 35/43 × RAT × Δy exv. fy = 35/43 / RAT × Δx where RAT is the aspect ratio with respect to the pixels of the imaging system to be used, and in this example, RAT = 3.47 / 2.70.

As described above, when the marker position is calculated by extrapolation, the extrapolation vector is calculated from the positional relationship of the previously detected marker pair, whereby the marker set forming the block corresponding to the marker pair is calculated. It can be detected reliably.

The extrapolated vector 104 can be used in the processing of the next frame.

That is, the entire dot code 36 is 1
The image is not always captured as a single screen. In practice, many frame images are continuously captured and imaged in many cases. In this case, the skew angle hardly changes suddenly between two temporally consecutive frames, so that the extrapolated vector 104 obtained in the previous frame can be used in the next frame.

As described above, if the extrapolation vector 104 of the immediately preceding frame is referred to in the second and subsequent frames, it is not necessary to obtain the extrapolation vector 104 by calculation. That is, it is possible to omit the processing of the marker presence area calculation unit 88.

Further, when searching for a corresponding marker in the marker search area 106 obtained by the extrapolation vector 104,
At that time, the actual marker may be slightly shifted from the extrapolated position due to, for example, distortion or inclination. Therefore, in such a case, as shown in FIG. 19B, the extrapolation vector is corrected using the approximate center of the paired marker 100 searched in the marker search area 106, and the next paired marker is corrected. At the time of searching 100, it is preferable to set the marker search area 106 by the corrected extrapolated vector, that is, the corrected extrapolated vector 108.

As described above, using the marker center position detected based on the extrapolated vector 104, this extrapolated vector 1
04, that is, by proceeding while correcting the extrapolation vector, the marker can be reliably detected without losing the marker even when there is distortion or inclination.

If all of the paired marker candidates are detected by any of the above methods, the marker center detecting section 56 uses the base marker pair selecting section 92 to convert a series of marker pairs extending over two blocks into base marker pairs. Select as a pair.

For example, as shown in FIG. 20A, when there is a series of marker pairs extending over two blocks, such as a marker pair including markers M10 and M11 and a marker pair including markers M11 and M13. , Those three markers M
10, M11 and M13 are selected as a base marker pair.

That is, the existence area of the adjacent marker, that is, the marker existence area 98 is calculated from the reference marker (ntm) based on the specification skew angle and the format information, and the marker candidates in the marker existence area 98 are detected. Are extracted as adjacent markers (lmc, rmc). Then, the reference marker and the extracted marker are selected as a base marker pair 110.

That is, when there is an erroneously detected marker, the middle marker becomes exactly the middle point when three markers 16 including the erroneously detected marker are connected due to the format limitation that they are arranged in a line. Therefore, if a marker pair extending over two blocks is selected as a base marker pair, such a false detection marker is rejected and only the actual marker is selected. Will be. That is, it is possible to reject a marker erroneously detected as a marker due to noise.

When the dot code 36 is recorded at double height, by selecting such a base marker pair extending over such two blocks,
Here you can decide whether to read the columns.

Further, since the image data spans a plurality of blocks, more pattern dots can be used and the accuracy of the data reading reference point can be increased. That is, in the example of the drawing, not only is the reading reference point determined using the pattern dots between the markers M10 and M11, but also the reading reference point, that is, the marker is determined using the pattern dots in the range from the markers M10 to M13. By determining the true center coordinates, the accuracy can be increased accordingly.

Also, since the data reading reference point can be determined simultaneously over a plurality of blocks,
The true center of the markers M10 and M11 is obtained, and then the marker M
There is also an advantage that the amount of calculation is smaller than in the case where the true center of 11 and M13 is obtained sequentially and so on.

As described above, if a series of marker pairs extending over at least two or more blocks are selected as the base marker pair, the format limitation of the base marker pair (the pair is limited) causes Erroneous marker detection due to noise can be reduced. Further, in the case of a code recorded in double height, by selecting a marker pair that straddles the reproduced block sequence as a base marker pair, double reading can be eliminated. Furthermore, by extending over a plurality of blocks, many pattern dots can be read, and the accuracy of the data reading reference point is increased. Further, since the data reading reference point can be determined simultaneously over a plurality of blocks, there is an effect that the amount of calculation is small.

Here, the double height will be described.

The double height means that exactly the same code is printed in two rows as shown in FIG. 20B (A1, A2,... Indicate block addresses in the figure). In other words, the original code has two rows (two vertical blocks)
In the case of overlapping, in this double height, there are four stages (four vertical blocks). When the dot code 36 is scanned in the horizontal direction by the image input unit 38,
As a frame image obtained by the image pickup unit 46 (image pickup unit visual field 112) such as a CCD camera, an image rotated by 90 ° is obtained as shown in FIG. That is,
As the scan is performed, a frame image in which the blocks are sequentially slid upward is obtained, such that the blocks rise from the bottom.

In this case, from the relationship between the size of the block and the size of the frame, three blocks are usually captured in one frame in the horizontal direction. Of these three blocks,
In the case of double height, since two blocks have the same data content, it is only necessary to perform processing on only one of the blocks. Therefore, in the example of FIG. 17A described above, the markers M5, M6, M10, M
8 in four corners and markers M7, M9, M1
Blocks M3 and M11 have the same data contents as the blocks arranged at the four corners. Since the former block is selected as the processing target block 82, the latter block is excluded.

The base marker pair selection section 92 selects a reference marker from the marker candidates in the scanning direction at least within a range where the left and right adjacent markers can be detected, and selects a base marker pair based on the reference marker.

That is, when a certain marker is found, the position of the marker adjacent thereto can be specified by the specification skew angle and format as described above. In this case, the reduction ratio due to the inclination is also related, but the adjacent marker always enters the frame 60 only when the marker is in the area 114 as shown in FIG. Therefore, this area 114
Select the reference marker ntm.

[0145] This area 114 is
The area is smaller by a distance ACRY from the upper and lower ends of the frame and by a distance ACRX2 from the left and right ends. Here, ACRY and ACRX2 are, as shown in FIG. 21B, a half of the length of the paired marker existence region 98 in the Y direction and a distance from the marker to the far end of the paired marker existence region 98. This value corresponds to the length in the X direction and is calculated by a mathematical formula as shown in FIG.

That is, if a reference marker is selected in such an area 114, no matter which marker in this area 114 is selected, there is a possibility that adjacent markers will always be detected on both sides.

The selection of the reference marker from the marker candidates in the scanning direction is for the selection of a block to be described later. When the blocks are selected from the scanning direction in this way, the reference frame is selected in the previous frame. Blocks that have already been processed will be selected later and come out,
If the processing of the frame is stopped when a processed block is detected, useless processing for the already processed block can be omitted.

In the illustrated example, the reference marker indicates the center marker of the base marker pair 110, and the lower marker of the frame is set as the reference marker ntm as much as possible. Is selected as the marker M11.

As described above, the reference marker is selected from the marker candidates in the scanning direction within at least the range in which the left and right adjacent markers can be detected, and the base marker pair is selected based on the reference marker. When the adjacent marker is not detected, it can be estimated that the code is off the screen or that an accident has occurred. Further, by selecting a reference marker from the marker candidates on the scanning direction side, processing is performed from an undetected marker in the previous frame, so that overlapping processing can be prevented.

When a marker candidate is found on only one side of the reference marker ntm, the base marker pair selection section 92 reselects the marker as a reference marker and searches for an adjacent marker. However, at the time of this marker search, the existence area is limited by extrapolation.

For example, in FIG. 21C, when the marker M10 is first selected as the reference marker ntm and the corresponding markers lmc and rmc are searched from the paired marker existence area 98, the marker M10 is selected from the right paired marker existence area 98. When the marker is not found, and the marker M9 is found as the left adjacent marker lmc from the left pair marker existing area 98, it is determined that the marker M10 is the marker at the end of the dot code 36. Therefore, the found marker M9 is used as the reference marker ntm. Then, the marker search area 1 is obtained from the already detected marker M10 by using an extrapolated vector.
06, and a left adjacent marker lmc is searched in the area. Thus, the marker M8 is detected, and the markers M8, M
9, M10 is a base marker pair.

As described above, when only one of the left and right markers can be detected with respect to the marker, more blocks can be read by replacing the reference marker with the detected adjacent marker and searching for the marker by extrapolation. . That is, since the block is read based on the reference marker, when the marker at the code end is selected as the reference marker, the number of blocks that can be read decreases. Therefore, if the selected reference marker is a code end, by reselecting it, more blocks can be read. Further, since the search area is narrowed by extrapolation, erroneous detection of a marker due to noise can be reduced.

When the extrapolated marker position protrudes from the frame 60 as described above, the base marker pair selecting unit 92 stores the extrapolated position as the approximate center of the virtual marker.

For example, in FIG. 22A, when the marker M8 is first set as the reference marker ntm and the right adjacent marker rmc is not found when searching for the paired marker existence area 98 on both sides, as described above. Next, the reference marker is moved to the marker M7, and the left adjacent marker lmc is searched for by extrapolation from the marker pair consisting of the markers M7 and M8. At this time, if the extrapolated point is out of the frame 60, a virtual marker is provided at the extrapolated point.

By doing so, the marker M in FIG.
The pattern dots 14 between the virtual marker 7 and the virtual marker M10 can be used.

When such a virtual marker M10 is not provided, only a set of markers M7 and M8 can be found, and only blocks in the same column as the block including the markers M7 and M8 are read. I will. However, by providing the virtual marker M10 in this manner, a block in which the markers M1, M2, M5, and M4 are arranged at the four corners can also be read, and it is possible to prevent a block from being lost.

As described above, when the marker search area protrudes from the screen when extrapolated, the extrapolation point is set as a virtual marker, so that even if the marker protrudes from the screen, the pattern dots remaining on the screen are displayed. Since the pattern dots 14 can be used, these pattern dots 14 can be effectively used.

The blocks are sequentially read out by extrapolation with reference to the reference marker. Even if a marker adjacent to the reference marker is off the screen, the marker constituting the block obtained by extrapolation is not skewed. May enter the frame, and in such a case, those blocks can be read.

When two or more base marker pairs can be selected, the base marker pair selecting section 92 reselects a base marker pair that has a high possibility of reading more blocks.

That is, in the case of the double height, as described above, exactly the same code is printed in two rows. As the frame image, when the original code is composed of two horizontal blocks, there are four horizontal blocks.
In general, three horizontal blocks are imaged in one frame because of the size of. Therefore, by re-selecting a base marker pair that is likely to be able to read more blocks,
The number of read blocks can be increased.

For example, when the markers M10, M11, and M13 are selected as a pair of base markers as shown on the left side of FIG. 22B, the read block is displayed as shaded on the right side of FIG. , Markers M2, M4, M7, M
6 at the four corners, M6, M7, M11, M
There are three blocks of ten blocks and blocks of M7, M9, M13, and M11.

On the other hand, as shown on the left side of FIG. 22C, when the markers M8, M10 and M11 are selected as a base marker pair, as shown by hatching on the right side of FIG. The read block includes the markers M1, M2, M
6, M5 at the four corners, and M5, M6, M1
0, M8 block, M8, M10, M14, M12 block, M2, M4, M7, M6 block, M6,
There are five blocks of blocks M7, M11 and M10.

Here, in the case of double height, M5,
Blocks M6, M10, M8 and M7, M9, M1
Since the same data is recorded in the blocks 3 and M11, the data obtained by reading either block is the same. Therefore, M8, M10, and M11 having the largest number of readable blocks are selected as a base marker pair.

As described above, when two or more base marker pairs can be selected, by re-selecting a base marker pair having a high possibility of reading more blocks,
More read blocks can be extracted from one frame.

When the base marker pair is reselected as described above, the right base marker pair is displayed when the code is skewed to the left, and the left base marker pair is displayed when the code is skewed to the right. When the marker pair is reselected and the marker is searched, the existence area is limited by extrapolation.

For example, when the code is skewed to the right as shown in FIG. 23A, a set of M8, M10, M11 on the left and M1 on the right are used as a base marker pair.
There is a possibility that a set of 0, M11, and M13 is selected. When the left set is selected, five blocks can be read, but when the right set is selected, only four blocks can be read. Thus, when the code is skewed to the right, the left marker pair is selected.

The markers constituting the block are sequentially obtained by extrapolation based on the base marker pair. As described above, when the base marker pair is reselected, if the code is skewed to the right, the left marker is selected. When the base marker pair is skewed to the left, by reselecting the right base marker pair, it is possible to prevent the markers constituting the read block from being out of the frame due to the skew and causing block loss ( Lateral direction).

Alternatively, in the reselection of the base marker pair, a base marker pair from the center may be selected. This is performed by reselecting a marker pair near the center when the midpoint between the left marker and the reference marker is on the right side of the frame center, or when the midpoint between the right marker and the reference marker is on the left side of the frame center. .

For example, as shown in FIG. 23B, when the code is skewed to the right, the markers M10, M1
When M1 and M13 are detected as a base marker pair, the midpoint between the left marker M10 and the reference marker M11 is the frame 60.
Is located on the right side of the center of the frame 60, it can be seen that this base marker pair is shifted to the right side of the frame 60. Therefore, in such a case, a marker pair consisting of M10 and M11 closer to the center than a marker pair consisting of M11 and M13 is reselected. Then, a marker M8 is searched by extrapolation, and M8,
M10 and M11 are selected as a base marker pair.

If the code is skewed to the left,
By determining whether the midpoint between the right marker and the reference marker is on the left side of the frame center, it can be determined whether or not to reselect.

The markers constituting the block are sequentially obtained by extrapolation with reference to the base marker pair. In this way, the midpoint between the left marker and the reference marker is on the right side of the frame center, or the right marker and the reference marker When the midpoint between the two is located on the left side of the center of the frame, by reselecting the marker pair near the center, it is possible to prevent the markers constituting the read block from protruding from the frame and causing block loss due to skew. In addition, since the block from the center is selected, the influence of distortion can be reduced.

Further, the marker set selecting section 94 selects all or a part of the markers constituting the block.

That is, as an example, as shown in FIG. 24A, when the block 20 has a rectangular shape and markers are arranged at its vertices, all the blocks constituting the block 20 are displayed.
One marker set is selected.

As described above, by selecting the marker set constituting the block 20, the data to be read can be divided and processed for each block.

The marker set selecting section 94 stores all the marker sets constituting the selected block 20 in a memory (not shown).

For example, in the case of the example of FIG.
Markers M0, M2, M6, M
When it is found that 4 constitutes one block, it is used as a marker set of the first block, as shown in FIG. 4B, M0 is at the upper left, M2 is at the upper right, M6 is at the lower right, and M6 is at the lower left. M4 is stored in the memory. Similarly, M1, M3, M7, and M5 are stored as the second block, M8, M10, M14, and M12 are stored as the third block, and M9, M11, M15, and M13 are stored as the fourth block. That is, since there are four blocks in this example, the first to fourth blocks are stored in the memory with the marker numbers in this manner.

As described above, the block 2 is determined based on the approximate center position.
By storing all the marker sets constituting 0 in the memory, data can be read for each block even when the blocks 20 are not aligned.

Alternatively, instead of storing all the marker sets, a predetermined one marker constituting the block 20 and a relational expression expressing the relation between this marker and another marker are stored in the memory. May be.

That is, when the processing target blocks 82 are shaded as shown in FIG.
If a predetermined one of the four markers constituting the block, for example, the upper left marker and the relational expression between the upper left marker and the remaining three markers are known, all the marker sets are stored in the memory. do not have to.

For example, in the case of the example shown in FIG.
As shown in the relational expression, if the upper right marker is stored as +1, the lower left marker is stored as +3, and the lower right marker is stored as +4, the upper left marker M is obtained for the first block.
Just store 1 and the upper right marker is M
2, lower left marker is M4 +3, lower right marker is +4
It can be seen that the marker M5 is
4, M5 need not be stored in the memory.

As described above, if the relationship between at least one of the markers constituting the block 20 and another marker is known, it is only necessary to store the at least one marker in the memory, and the memory capacity can be saved.

Further, as shown in FIG. 25A, the marker set selecting section 94 outputs all the markers constituting the block 20 so that all the blocks 20 included in the frame 60 can be processed. A set may be selected. However, in this case, the base marker pair selection unit 9
There is no need to configure 2.

In the case of double height, exactly the same code is printed in two rows as described above. As a frame image, when the original code is composed of two horizontal blocks, the horizontal four
Generally, three horizontal blocks are imaged in one frame from the relationship between the block size and the size of the frame 60. Therefore, as shown in FIG.
When all the marker sets constituting 0 are selected, blocks having the same content are read in duplicate, and the data read in the same manner in both cases is regarded as correct data, thereby improving the data reliability.

That is, the reliability can be improved by treating all the blocks selected redundantly as processing targets.

Further, as shown in FIG. 25B, the marker set selecting section 94 sets the marker constituting the block in the scan direction so that the block in the scan direction can be changed to the processing target block 82 in the frame. Select from the set in order, and if a processed block appears in the previous frame,
Therefore, the processing may be stopped and the processing may be shifted to the next frame.

For example, when the reference marker is selected from the marker candidates in the scanning direction in the base marker pair selection unit 92 as described above, as shown in FIG.
Markers M8, M10, and M11 are selected as a base marker pair, but in order to read from the lower block, a process of first extrapolating downward once to find a marker is performed. As a result, the markers M12 and M14 are detected as markers, and these are regarded as a set of markers constituting one block.
Blocks 0, M14 and M12 are selected. Next, on the contrary, the upper block is selected and processed. Then, since it is known from the block address whether or not the block has already been processed, when the already processed block is found, the processing for this frame is stopped, so that the already processed block is not processed again. Can be omitted.

As described above, by sequentially selecting from the marker set constituting the block on the scanning direction side of the frame, the processed block can be excluded from the processing target, so that the redundant processing can be omitted.

If the image pickup section 46 has an optical distortion or the like, the distortion is larger on the outer side than on the center of the frame 60. If the data of the block closer to the center is read first, then if the data of the block closer to the center can be read, the data is overwritten with that data. Can be made more reliable.

For example, when a code is recorded in double height, as shown by a circle in FIG.
The data of the processing target block 82 is read in the order of .about.9. Here, the same data is recorded in the first block and the second block, but since the one block is located near the end of the frame 60, there is a high probability that a reading error will occur due to optical distortion. Then, if the data of the second block is read once again and overwritten, more accurate data can be obtained. Also, for example, even if the two blocks cannot be read due to noise or the like, the data of this block may have some errors since the data of the block is read from the first block first. It can be prevented from being read.

In other words, the closer the block is to the outside of the frame, the greater the effect of distortion and the more errors occur. Thus, in this manner, the marker set constituting the block near the end of the frame in the direction orthogonal to the scan direction is sequentially set. By making selection, if data in a block is sequentially read from the outside and overwritten, highly reliable data can be obtained without requiring an extra memory.

Alternatively, when the reading of the selected block fails, the marker set selecting section 94 may reselect a block having the same address.

For example, as shown in FIG. 27A, in the double-height code, the middle three markers (not shown) are selected as a base marker pair, and the processing target block 82 is read in order. In the case, the address A17
Becomes a read failure block 116, the marker set is selected again so that a block having the same data (the same address) as this is selected as the reselection block 118.

A read error occurs and the read failure block 116 occurs, for example, when the address dot 18 cannot be read due to noise or when a marker cannot be found.

As described above, when the reading of a certain block has failed, by reselecting the marker set constituting another block having the same address, it is possible to reduce the reading error due to the drop of the block.

Next, a method of estimating the reading reference point of the data code 10 will be described.

As shown in FIG. 27B, the marker 16
And the pattern dots 14 are arranged on the code line 120 in a straight line. That is, the true center of the marker 16 as a reading reference point of the data code 10 and the center of each dot of the pattern dot 14 are arranged on the code line 120. Therefore, by using the pattern dots 14 recorded in the predetermined format, it can be immediately determined whether or not the approximate center of the read marker is the correct true center.

Further, by using a plurality of pattern dots 14, the accuracy of the read reference point to be obtained can be improved. That is, since the format of the pattern dot 14 is determined, when the pattern dot 14 is detected, it can be discriminated whether the dot is really at a position where the dot should be or is a noise. Therefore, the center of all the dots of the pattern dot 14 at the desired position can be obtained, and the true center of the marker 16 can be obtained using the center positions of the plurality of dots. The accuracy of the position of the reading reference point can be improved.

For example, as shown in FIG. 27C, assuming that the pattern shape of each dot of the pattern dot 14 after the binarization processing is obtained, the read line 122 connecting the approximate centers of the markers 16 is obtained. Is detected at the dot reading position 124 according to the format information. Then, the center position 126 of the detected dot is calculated, and the true center of the marker serving as the reference point for reading the data code 10 can be calculated from these coordinates by linear regression.

The accuracy of detecting the center position for each dot is approximately several pixels of an error, but the accuracy can be improved by preparing a plurality of pattern dots 14 for one marker 16.

As described above, when the data reading reference point is estimated by using the pattern dots 14 recorded in a predetermined format, it can be immediately determined whether or not the read information is correct by comparing with the predetermined format. Also, at least one or more pattern dots 1
4, the reference point estimation accuracy can be improved.

When the data reading reference point, that is, the true center of the marker 16 is estimated using the pattern dots 14 as described above, accurate data is obtained by using only the pattern dots 14 having a predetermined positional relationship from the approximate center of the marker. Estimate the reading reference point.

The pattern dots 14 having the predetermined positional relationship are, for example, pattern dots 14 between adjacent markers as shown in FIG. The estimation of the true center of the marker is performed one block at a time to deal with distortion. For example, when finding the true centers of the markers M8 and M10, the pattern dots 14 between them are used. When finding the true centers of the markers M10 and M11, the pattern dots 14 between them are used. Used. In this case, the true center of the center marker M10 is obtained from the pattern dots 14 on both sides, respectively.
(B), the average of the left and right blocks is used.

Alternatively, as shown in FIG. 28C, the pattern dots 14 having the predetermined positional relationship may be pattern dots 14 within a predetermined range including the pattern dots 14 outside the block. In this example, the true centers of M8, M10, and M11 are L, M,
It is determined using pattern dots in the range of R.

In this case, for example, if there is no optical distortion with respect to one data reading reference point to be obtained, that is, the true center of the marker 16, the pattern dots 14
Since the accuracy is better when the number of dots is calculated using a large number of dots, if there is no such optical distortion, as shown in FIG. It is preferable to find the true center.

In contrast, when there is optical distortion,
Since errors are accumulated as the distance from the marker 16 increases and a large error occurs, the range of the pattern dots 14 to be read is limited as shown in FIG. By using only the neighboring pattern dots 14), the reliability of the obtained data reading reference point is increased.

As described above, by estimating an accurate data reading reference point using only the pattern dots 14 having a predetermined positional relationship from the approximate center of the marker, if there is no distortion, the range is widened and the number of dots is reduced. By increasing the number, the estimation accuracy can be increased. When there is distortion, estimation accuracy can be improved by limiting the range. Furthermore, there is an advantage that the processing is simplified by setting the range between the marker pairs.

When the pattern dots 14 between the markers are used, a data reading reference point can be estimated by employing a widely used tree search technique. That is, as shown in FIG. 29C, not only the pattern dots 14 are read on a line connecting the approximate centers of the marker pairs, but also the pattern dots 14 are read on a line connecting each of the eight approximate centers. . That is, reading is performed nine times for each point. In such a case, not all the dots of the pattern dot 14 can be read at the shifted position, and all of the dots can be read at the correct position. Therefore, as shown in FIG. The most frequent point (indicated by a circle in the figure) is defined as a data reading reference point. Alternatively, a more accurate reading reference point obtained by similarly performing reading using eight points in a narrower range near the reference point obtained as described above may be used.

As described above, when the tree search method is employed, the data reading reference point can be determined without requiring complicated calculations.

Further, in this case, instead of performing a brute force tree search, the pattern dots 14 may be read once between the approximate centers of the marker pairs, and the tree search may be limitedly performed according to the result.

For example, as shown in FIG. 30A, when dots 1 and 10 are set at the approximate center of the marker, dots 2 and
When 3, 4, and 5 are read and dots 6, 7, 8, and 9 are not read, that is, when only the dots on the left are correctly read, the approximate center (dot 1) on the left is almost correct. Position and the approximate center on the right (dot 1
0) can be determined to be shifted. Therefore, the tree search may be performed while leaving the approximate center on the left side as it is, and moving only the approximate center on the right side to around eight.

Further, as shown in FIG. 30B, the pattern dots 14 are shifted to the right by the dots 2, 6, the dots 3, 5, 7 at the normal position, and the dots 4, 4 shifted to the left. In the case where only the dots 4 and 8 shifted to the left cannot be read when this is read, it is understood that the approximate center position is shifted to the right. That is, the fact that the reading error has occurred in the dots 4 and 8 shifted to the left means that the dots 4 and 8 are generally shifted to the right. In such a case, the approximate centers of the dots 1 and 9 are shifted. When the tree search is continued, the points shifted to the left thereof may be mainly searched.

Similarly, by inserting dots vertically displaced into the dot pattern 14 as shown in FIG. 7B, the displacement in the vertical direction can be detected, and the tree search is accordingly performed. Can be limitedly performed.

Then, when all the dots of the pattern dots 14 are detected almost uniformly, it means that the data reading reference point has been detected almost surely.

As described above, the pattern dot is read at the dot read point determined by the approximate center of the marker pair and the format, and the amount and direction of the shift of the approximate center position are estimated based on the position of the dot where the read error has occurred and its bias. By performing the tree search by moving the approximate center of the marker pair in the vicinity based on the estimation result, the number of searches can be reduced as compared with the brute force tree search.

Note that the pattern dots 14 do not necessarily have to be linearly arranged between the marker pairs, but it is only necessary to determine the positional relationship with the marker pairs as a format.

For example, as shown in FIG. 30C, when the pattern dots 14 are arranged in a circle between or around markers (not shown) and the center thereof is used as a data reading reference point, r for pattern dot 1
In the case of the format in which 4 are arranged, each dot is detected from the approximate center of the marker, and the center (indicated by x in the figure) is obtained. And the center (x, y) of the circle of radius r
Is determined so that the sum of squares of the distance between the center of the eight dots and the circle is minimized, and is used as a data reading reference point. This means that in image terms, the radius r
Is moved to find a place where the calculated center of the dot fits into this circle, and if found, the center (x, y) of the circle is used as the data reading reference point. ing. More simply, the center of gravity of these dots may be set as the center of a circle, that is, a data reading reference point.

As described above, the pattern dot is detected at the dot read point determined by the approximate center of the marker pair and the format, and the center of the detected pattern dot is calculated. Since the reading reference point is estimated so as to minimize the predetermined error function defined by the relationship, the data reading reference point is estimated using the centers of the plurality of pattern dots, and the accuracy is extremely high. Become.

As described above, in the marker center detecting section 56, as shown in FIG.
, The pattern dots 14 arranged between the markers of the marker set selected by the marker set selecting unit 94 are detected at the approximate center of each marker and at the dot reading point determined by the format, and the dot center calculating unit 128 To calculate the center of each detected dot. Then, the error minimizing unit 13
Based on 0, a reading reference point of each data dot of the data code 10 is estimated so as to minimize a predetermined error function defined by the relationship between the calculated dot center and the format, and outputs it to the data reading unit 42.

Next, the pattern dot detecting section 126 will be described in detail.

The dot detecting section 126 has a black dot detecting section 132 for detecting a black dot at a dot read point determined by the approximate center of the marker pair and the format. As shown in FIG. 31 (A), the black dot detection unit 132 detects the dot reading point 13 where a black dot should be detected.
If a black dot is not detected in step 4, a black pixel is searched for four neighboring pixels in the order indicated by 1 to 4 in FIG. Explore. That is, each dot forming the pattern dot 14 is detected as a pixel value of a dot reading point satisfying the predetermined positional relationship from the approximate center of the marker. Therefore, when a black dot is not detected at a dot reading point at which a black dot is to be detected, it is considered that the dot reading point has shifted, and the dot is searched for while shifting one pixel at a time up, down, left and right.

As described above, even if the approximate center of the marker is calculated by shifting due to noise or the like, or the dot reading point 134 obtained by dividing the approximate center by a predetermined ratio due to distortion or inclination is calculated by shifting, By searching for black dots in the vicinity, the pattern dots 14 can be reliably detected, so that there is no decrease in accuracy due to a decrease in the number of detected dots.

Further, the dot reading point 1
If a dot is detected in the vicinity pixel of 34, the shift amount is stored, and is stored in the next dot reading point 1
Preferably, it is added as an offset from 34.

That is, for example, in FIG. 31B, if the dot indicated by the black circle on the left side in FIG. 31 is the dot reading point 134 obtained by calculation, there is no black pixel there, so a search is performed. When a dot is detected at a position shifted by one pixel to the left, a position shifted by one pixel to the left is stored as an offset. Then, even if the next dot reading point 134 is a point indicated by a white circle in the figure in the initial calculation, a point shifted by one pixel to the left from that point is read as the dot reading point 134.

In other words, when a certain dot is detected with a shift, it is very likely that a dot in the vicinity is also detected with a shift. If the point to which the offset is added from the beginning is read, the processing is reduced.

As described above, when a certain dot of the pattern dot 14 is detected at a shifted position, the dots near the same are often shifted in the same manner. When a dot is detected at a shifted position, the search efficiency can be increased by shifting the next dot reading point 134 in advance using the shift as an offset. Also, if the dot detection point is shifted due to the distortion, unless the dot reading point 134 is shifted in accordance with the amount of distortion, the deviation will accumulate and eventually the dot detection becomes impossible. By doing so, this can be prevented.

As shown in FIG. 18, the marker center detecting section 56 further calculates a marker center center based on the data reading reference point obtained by the error minimizing section 130 to recalculate the marker center. May be provided.

That is, each dot forming the pattern dot 14 is detected as a pixel value of the dot reading point 134 satisfying a predetermined positional relationship from the approximate center of the marker. Therefore, when only a few dots are detected, it is assumed that the approximate center of the marker has shifted, and the approximate center position of the marker is recalculated using only the dots that can be read. The dot is detected by reading the pixel value of the dot reading point 134 satisfying the predetermined positional relationship again from the approximate center of the marker thus obtained.

For example, the approximate center of the marker is detected as indicated by a cross on the upper side of FIG.
The first reading of No. 4 is performed. At this time, if the right center marker is misaligned and detected, the dots closer to the left marker approximate center can be read, but the dots closer to the right marker approximate center are missing. Would. When the center position of the dot read in this way is calculated by the dot center calculation unit 128 and the error minimizing unit 130 estimates the data reading reference point, the estimated data reading reference point has few detected dots. Therefore, there are many errors. Therefore, the marker approximate center recalculation unit 136 sets the data reading reference point as a new marker approximate center, and sets
26 causes the pattern dots 14 to be detected again. As a result, as shown in the lower part of FIG.
Are detected, and the data reading reference point can be estimated with high accuracy.

As described above, by re-calculating the approximate center from several pattern dots detected in the first reading, the accuracy of the approximate center is improved, and the number of re-detected dots can be increased.

Further, since the black dot detecting section 132 only detects black dots, a point which was originally white due to dirt or the like may be detected as a black dot. Therefore, it is also possible to configure the pattern dot 14 by a signal + error detection code, perform error detection processing from the detected black dot signal, and calculate the dot center only at the point that was originally a black dot. .

That is, as shown in FIG. 18, in the pattern dot detecting section 126, the detection result of the black dot detecting section 132 is input to the buffer 138, and is arranged as a signal. That is, assuming that a black dot is “1”, for example, 1, 0, 0,
If rearranged as 1, 1, 0, 0, this is a signal, so that error detection becomes possible. Therefore, the error detection unit 140 performs an error detection, and if there is a black dot that is erroneously detected, the selection unit 142 uses the black dot detection so that it is not used, that is, is not supplied to the dot center calculation unit 128. The output of the unit 132 is selectively supplied to the dot center calculation unit 128.

As described above, after the dots are detected, by selecting only correct dots from the detected dots by error detection, the dots are correctly detected even when the pattern dots 14 have information. Since the position is known, it is possible to prevent a decrease in data reading reference point accuracy due to noise.

Next, the dot center calculator 128 will be described.

When dots printed in a circle are roughly imaged by a CCD camera and binarized, the obtained pattern is no longer circular, so that the dot center cannot be easily calculated.

In the case where ideal binarization is performed, as shown in FIG. 32A, if half of a pixel touches a dot, the pixel turns on, and a dot for turning on a certain pixel turns on. The existence range of the center position can be defined. That is, when the trajectory of the dot center when half of the pixel touches the dot is defined as a critical line, the pixel is turned on when the dot center is inside (the target pixel side), and turned off when it is outside. Become.
Therefore, when a certain pattern occurs, if the existence range of the dot center position is determined for the turned-on pixel in the pattern and the surrounding off-pixels, the dot center position can be estimated as the overlapping area. .

For example, when one dot is imaged with a size of about two vertical pixels and three horizontal pixels, when ideal binarization is performed, the pixel pattern that appears is indicated by a circle in FIG. There are only seven types as indicated by numerals 1 to 7, and the area where the dot center position exists can be estimated as shown in each figure.

The dot center calculation section 128 holds the relationship between such a pattern shape and the dot center existence area as a table.

As described above, the existence area of the center position can be obtained from the pattern shape when the dot whose shape is known is binarized. Therefore, by setting the point in this area as the center representative point, It becomes possible to estimate the dot center in a super-resolution manner even with coarse sampling.

On the other hand, as described above, one dot is divided into two pixels × 3
When imaging with pixels, there are only seven patterns,
In the case of imaging with a higher resolution of 3 pixels × 3 pixels or 4 pixels × 3 pixels, the pattern appearing increases exponentially as the resolution increases, and the table is held as described above. Requires a very large amount of memory.

Therefore, the dot center calculating section 128 may calculate the center of gravity of the imaged dot and use the calculated center as the dot center. This eliminates the need for a memory because it is not necessary to hold the relationship between the dot shape and the representative point in a table. In addition, since it is not necessary to decide how many pixels × how many pixels to image one dot, it has a scalability that can cope with a change in the magnification of the optical system or a change in the diameter of the dot. There is also an effect.

When the center of gravity is calculated, as shown in FIG. 33A, a predetermined size, for example, a 4 × 4 mask 144 is applied to the dot read point 134, and
The center of gravity 146 may be calculated using only the pixels contained therein.

Here, the calculation of the center of gravity will be described with reference to the example of FIG.

First, in the same figure, there are 11 black pixels in the mask, and when these are viewed in the row direction, one black pixel is located in the row located one row above the row of the dot reading point 134.
Since there are pixels, the row weight coefficient “−1” of this row is multiplied by “1 (pixel)” and temporarily stored. Similarly, since there are four black pixels in the row located one row below the row of the dot reading point 134, the row weighting coefficients “1” and “4” are multiplied.
It is added to the temporary storage result described above. Further, the row below the second row is similarly multiplied by the row weighting factor “2” to the three black pixels, and the result “6” is added to the temporary storage result. The addition result thus obtained is divided by the number of black pixels to obtain “9/1
1 "is obtained, which corresponds to the displacement of the center of gravity 146 from the reading point 134 in the y direction.

The same calculation is performed for the x direction.
"(-1 × 3 + 1 × 3 + 2 × 1) / 11 = 2/11"
Is a value corresponding to the displacement of the center of gravity 146 from the reading point 134 in the x direction.

As described above, the mask 144 having an appropriate size is used.
If the region contributing to the center calculation is limited by the above, the influence of the error due to noise (stain) on the estimation of the reading reference point can be reduced.

In this case, in order to reduce the influence of noise, the size of the mask 144 is preferably as close as possible to the size of the dot. However, the size of the mask 144 is preferably the same as the size of the dot. If the dot reading point 134 is not detected at a position close to the center of gravity, a mask is applied to the biased position, and pixels contributing to the calculation of the center of gravity are also biased. Will be bigger.

Therefore, as shown in FIG. 33B, first, the mask 1 is set at a position with the dot reading point 134 as a reference.
44 to detect the bias of the dots in the mask,
As shown in FIG. 3 (C), if the mask 144 is moved in a direction to eliminate the bias before calculating the center of gravity, accurate calculation of the center of gravity can be performed even if the size of the mask 144 is adjusted to the size of the dot. Will be possible.

Here, the amount of movement of the mask 144 in the horizontal (vertical) direction is “○” if at least one black pixel exists in a column (row), and is “x” if no black pixel exists in FIG. (D)
The calculation is performed with reference to the table shown in FIG.

The following is a specific description with reference to FIGS. 33 (B), (C) and (D).

FIG. 33B shows an example in which the reading point 134 is shifted to the lower left, and nine black pixels are included in the mask.

First, in the row direction, the reading point 134
One row above, one row below, and two rows below,
Three and zero black pixels are detected. In this case, since there is no black pixel in the row two rows below, the mask position is shifted up one row.

Similarly, in the column direction, 0, 3
And three black pixels are detected. At this time, the mask 144 is shifted to the right by one column.

FIG. 33D summarizes these relationships in a table.

As described above, the mask 144 is applied with the detected position of the dot as a reference, the deviation of the dot is detected in the mask, the mask 144 is moved in a direction to eliminate the deviation, and the center is calculated. By determining the position in two passes, the mask size can be reduced in accordance with the dot size, so that less memory is required and errors due to noise are further reduced.

In the case where the existence area at the center position is obtained from the pattern shape obtained by binarizing a dot having a known shape as described above, the size of the center existence area greatly differs depending on the pattern. Therefore, as shown in FIG. 32C, the center representative point may be weighted according to the size of the center existing region, and the certainty may be reflected in the marker true center detection.

For example, in the case of the pattern indicated by the circled number 3 in FIG. 32B, the dot center existing area is a very small area, and thus has very high reliability. In the case of the pattern indicated by the numeral 4, the reliability is extremely low because this area is very large. Therefore, a weight is assigned to each area corresponding to the area of the center existence area. For example, if the weight of the center representative point in the case of the pattern indicated by the circle number 4 is “1”, “1” is “2.5”, “2” is “1.5”, and “3” is “3”. 5.0, "1.3" for the 5 pattern, "1.3" for the 6 pattern, and "1.5" for the 7 pattern.
And weight. By weighting the central representative point in this way, the reliability of the required data reading reference point can be calculated with reliability corresponding to the shape of this dot.

As described above, the estimation accuracy can be improved by estimating the data reading reference point by weighting the reliability of the central representative point which differs depending on the detected dot pattern shape.

Further, as shown in FIG. 34A, the error minimizing unit 130 calculates the center point of the dot actually calculated by the dot center detecting unit 128 based on the approximate center of the marker as described above. The center of gravity representative point position 150) and the estimated data reading reference point (in this case, the marker true center 1)
52) and the data reading reference point is determined so that the error between the estimated center point of the dot determined based on the format is minimized. In this case, using the least square method, The data reading reference point can be calculated with high accuracy.

Here, a method of calculating the data reading reference point (in this case, the true center of the marker) using the least square method will be briefly described.

Now, as shown in FIG. 34B, M1 ′
Is the true center of the left estimated marker, M2 'is the true center of the right estimated marker, Di is the i-th dot center of the pattern dot 14,
Let di be the inter-marker division ratio of the i-th dot of the pattern dot 14.

The dot center Di ′ obtained from the estimated marker true centers M1 ′ and M2 ′ is

(Equation 2)

Estimated error εi from dot pattern center Di
And the sum of squares E of the error is

(Equation 3) It is expressed as

Therefore, M1 ′, M2 that minimizes E
Ask for '.

[0264]

(Equation 4)

Here,

(Equation 5) In other words, from (1) Ax1 ′ + Bx2 ′ = P (3) From (2) Bx1 ′ + Cx2 ′ = Q (4) (3) × C− (4) × B

(Equation 6) From (4) × A- (3) × B

(Equation 7)

In the same way

(Equation 8) After all,

[0267]

(Equation 9) The position coordinates (x1 ', y1') and (x2 ', y2') of the estimated marker true centers M1 'and M2' are calculated.

As described above, the uncertainty of the calculated center position of each pattern dot is canceled by minimizing the sum of the squares of the shift amounts by the least square method, and the estimation accuracy of the data reading reference point is improved. be able to.

Further, as a method of simplifying the mathematical expression, there is a method described below.

Now, as shown in FIG. 35A, in the coordinate system on the recording medium, Ddi (ui) is the designed dot position, U is the first unit vector, O is the origin position, and As shown in FIG. 7B, in the coordinate system on the imaging surface, Doi (xi, yi) is the observed dot position,
(Ux ', uy') is the first estimated unit vector, and O '(o
x ', oy') is the estimated origin position, Doi '(xi, yi)
Is the estimated dot position.

Here, the designed dot position, the estimated reference point position, and the estimated dot position Doi ′ obtained from the unit vector are given by the following equations.

[0272]

(Equation 10)

Using the estimated error εi from the observed dot position Doi, the sum of squares E of the error is

[Equation 11] It is expressed as

Therefore, U ′ and O ′ that minimize E are obtained.

[0275]

(Equation 12)

Here, since U ′ and O ′ are obtained independently,

(Equation 13) Then, the expressions (5) and (6) are as follows.

[0277]

[Equation 14]

Similarly,

(Equation 15)

As described above, U ′ and O ′ can be obtained with a reduced number of calculations.

Here, an example in which a marker pair and ten pattern dots are arranged in a relationship as shown in FIG. 36 will be described.

First, the dot reading point Ui of each pattern dot is 7, 10, 1 from the approximate center of the left marker.
3, 16, 19, 24, 27, 30, 33, 36, and N = 10,

(Equation 16) Is 21.5.

However, if the second dot from the left (Ui = 10) is not detected by the pattern dot detection unit 126 due to a missing or dirty dot, N is 9 and Ui is 7,13. , 16,19,2
4, 27, 30, 33, and 36. Therefore, in this case, since N = 9, the solution of the above equation (11) is 22.7.
It becomes 8.

That is, when all of the 10 dots detected by the pattern dot detecting section 126 are detected,
Seen in the denominator of equations (7) and (9) above

[Equation 17] Is fixed at 902.5, but when some dots are not detected, the calculation is required each time.

On the other hand, the present dot code is often printed on a medium such as paper, and it is conceivable that dots may be missing depending on the printing conditions or the handling after printing.
There is a possibility that all zero dots are not detected.

Therefore, means for calculating the above equation (12) is required in the apparatus.
Since this is a process of adding the value of the square, the value of each term is large, and the number of bits of the register must be considerably large in consideration of the calculation accuracy. The scale also becomes large.

Therefore, when there is a dot not detected by the dot detection, the dot read point is set as the dot center of the dot. Then, processing is performed assuming that all dots have been detected.

By doing so, N = 10 and the solution of the above equation (12) is also 902.5.
All of the denominators in the above equations (7) to (10) can be fixed values.

That is, not only is there no need for a means for performing the calculation of the expression (12), but also the division of the expressions (7) to (10) is division by a constant, which can be replaced by multiplication of the reciprocal. It will be suitable for wear.

Hereinafter, this will be described with reference to the flowchart in FIG.

First, the reading points of the ten pattern dots are calculated from the approximate center of the marker pair (step ST1). Here, not only black dots but also white pixels between them are added as reading points. And
The dot detection counter n and the detection error counter err are reset to "0" (step ST2).

Next, each of the 30 reading points is sequentially processed, and it is determined whether or not the point is a point on a black dot (step ST3). Is determined (step ST4). If black, non-dot points are read as black, so the detection error counter err is counted as incorrect and the count is incremented (step ST5). If the pixel is white, the count is not incremented. Then, the number of read pixels is incremented, that is, the dot detection counter n is counted up (step ST6), and it is determined whether or not all the pixels have been read based on whether or not the dot detection counter n is "30". (Step ST7). If the number of pixels to be processed is less than 30, the process returns to step ST3 to process the next reading point.

On the other hand, when it is determined in step ST3 that a pixel on a black dot has been read, it is determined whether or not the dot has been read correctly by determining whether the read pixel is black (step ST3).
8). Here, when the dots are correctly read, the center of gravity of the connected pixel is directly calculated (step ST9).

If it is determined in step ST8 that the dot has not been read correctly, a black pixel is searched for in the four neighboring pixels of the reading point (step ST10).
Proceeding to 9, the calculation of the connected pixel is performed. If no dot is found, the dot read point is set as the dot center (step ST12), and the detection error counter err is counted up (step ST13). Then, the process proceeds to step ST6, where the dot detection counter n for counting the number of read pixels is counted up, and in step ST7, whether or not all the pixels have been read is determined by the dot detection counter n being “30”. Judgment is made according to whether or not. If the number of pixels to be processed is less than 30, the process returns to step ST3 to process the next reading point.

When reading of all pixels is completed in this way, the number of cases where non-dot pixels are black,
The total value of the dot position and the number of black pixels not present in the vicinity of the dot position 4, that is, the value of the detection error counter err is compared with a threshold value (MAX_ERR) (step ST14).
Incidentally, it has been found that the threshold value MAX_ERR is appropriately about 25.

If the value of the detection counter err is equal to or smaller than the threshold value, the data reading reference point is calculated by the least squares error method using the center of 10 dots (step ST
15) If the threshold value is exceeded, it is determined that the data reading reference point cannot be detected with the desired accuracy with the pattern code, and the process proceeds to the next marker pair.

When the data reading reference point is calculated as described above, the data reading unit 42
Based on this data reading reference point,
First, the address dot 18 is read by the address dot reading unit 154 to obtain a block address, and the data dot of the block 20 is read by the data reading unit 156. Then, the read data is supplied to the reproducing unit 44 and reproduced and output.

Here, the address dot reading section 154
Are the address dot reading unit 158 and the error correction unit 1
60, a determination unit 162, and an address determination unit 164.

The address reading section 158 divides the true center between the upper and lower adjacent markers to detect the address dot 18 and reads the address. If the read address is the address of a block that has already been read. Stops the processing and shifts the processing to the next frame. It should be noted that even when reading the address dots,
Similar to the above-described reading of the pattern dots 14, a tree search technique can be applied.

When the address dot 18 has a CRC (error detection) code, a BCH (error correction) code, or the like, the error correction unit 160 determines whether the address dot reading unit 158 has read the correct address. Judge and correct the read address.

The determining section 162 determines whether or not the address is within the address range of the data handled by the system. For example, in a system that handles only data from addresses 0 to 100, addresses such as -5 and 200 are excluded. Alternatively, if it is known that only 10 blocks are included in one frame and addresses are recorded in order, for example, addresses 2, 3, 5,
When 6, 25 and 9 are obtained, the address 2
5 is out of the range, and the data of the block at the address 25 is excluded.

[0301] The address determination section 164 gives the address determined as a correct address by the determination section 162 to the data reading section 156.

[0302] When an error is detected in the address dot 18 read by the address dot reading unit 158, the error correction unit 160 can calculate and correct the address of that block from the address of a surrounding block. However, by using this function, the address dot reading unit 158 may read the address with the address omitted as appropriate and determine the addresses of at least two or more blocks at the same time. That is, when the blocks are recorded in the order of the addresses, the address of a certain block can be easily estimated from the addresses of the surrounding blocks. The address dot reading process can be reduced accordingly. Furthermore, as shown in FIG. 9C, the address dots 18 can be recorded with thinning.

By the way, each block 2 forming the code
As shown in FIG. 38A, the address of the block is recorded in the left address dot 18 of 0.
The address of the next block is recorded in the rightmost address code of the code recorded with the double height as described above. Therefore, the last block 166
The address of the block is recorded in the address dot 18 on the right side of, so that the last block 166 can be detected by comparing the contents of the left and right addresses of the block 20 at the time of reading.

Alternatively, as shown in FIG.
A relative relationship indicating the last block address is provided between the addresses of the read block pair, and the last block can be detected by comparing the contents of the addresses of the block pair during reading.

For example, when recording 68 blocks in double height, if addresses 1 to 34 are arranged in a line and addresses 35 to 68 are arranged in a line, and these are recorded in double, that is, double height, the data is first taken in. Since the blocks at addresses 1 and 35 are detected adjacent to each other from the frame image to be read, one row is composed of 35-1 or 34 blocks. You can see immediately. Therefore, the amount of memory to be secured and the number of blocks to be processed before finishing the processing can be determined first, so that there is an effect that unnecessary processing is eliminated.

Of course, these are the same even when they are not recorded at the double height.

As described above, the format of the address dots 18 is a code in which the address of an adjacent block has a predetermined relationship with the number of blocks of the entire code, and the total number of blocks is obtained from the adjacent address. Memory can be allocated. In addition, since the last block can be detected, unnecessary processing can be omitted.

Alternatively, the address dots 18 may be recorded in a format in which the total number of blocks is known, for example, 1/68.

On the other hand, the data reading section 156 is to equally divide each side of an area surrounded by connecting four data reading reference points, cut a mesh, and read data codes as data reading points. In this case, as shown in FIG. 38C, the pixel at the data reading point 168 may be read differently from the actual data due to noise or the like.

Therefore, the data at the data read point 168 is not used as read data as it is, but pixels around eight points around the data read point 168 are used, and a predetermined calculation formula is used for all nine pixels, for example, The data of this pixel is determined by obtaining an average. For example, even if the pixel at the data reading point 168 is blank, if the surrounding eight pixels are black, the data reading point 16
The data of 8 is 8/9, which is larger than 0.5, so that it can be read as a black pixel, that is, "1".

The predetermined calculation formula may be a weighted average or the like. For example, a slightly larger area, that is, 24 pixels nearby, may be multiplied by 2 for the data of the inner 8 pixels, and multiplied by 1 for the data of the outer 16 pixels, and the average may be calculated. .

As described above, the vicinity of the reading point calculated from the data reading reference point is also read at the same time, and the attribute of the data reading point is determined by a predetermined calculation formula.

The present invention has been described based on the embodiments. However, the present invention is not limited to the above-described embodiments, and various modifications and applications are possible within the scope of the present invention. .

[0314]

As described in detail above, according to the present invention,
Since a block having a small influence of optical distortion is selected and read from a plurality of blocks in a captured image, the probability of occurrence of a reading error can be reduced. Therefore, it is possible to provide an information reproducing system capable of reading a code pattern recorded at high density and reproducing information correctly.

[Brief description of the drawings]

FIGS. 1A to 1C are diagrams showing examples of two-dimensional codes (dot codes) recorded optically readable on an information recording medium.

FIGS. 2A to 2E are diagrams showing examples of marker arrangement when blocks are formed as shown in FIG. 1C;

FIGS. 3A and 3B are diagrams illustrating code examples when markers are omitted in stripes, respectively.

FIG. 4 is a diagram illustrating a code example when markers are omitted in a checkered pattern;

FIGS. 5A to 5C are diagrams showing examples of pattern dots, respectively.

FIG. 6A is a diagram for explaining pattern dot modulation, and FIG. 6B is a diagram showing another example of a pattern dot.

FIGS. 7A and 7B are diagrams showing examples in which recording points of some pattern dots are provided with a phase shift.

FIGS. 8A and 8B are diagrams showing examples of address dots, and FIGS. 8C and 8D are diagrams showing different recording examples.

FIGS. 9A to 9C are diagrams showing still another example of recording address dots.

FIG. 10 is a block diagram of an information reproducing system according to an embodiment.

FIGS. 11A to 11D are diagrams for explaining an erosion process; FIGS.

FIG. 12 is a diagram for explaining an erosion process and a labeling process.

FIGS. 13A to 13C are diagrams for explaining a method of extracting a marker center part.

14A is a diagram for explaining a streak process, and FIG. 14B is a diagram for explaining another method of a marker extraction process.

FIG. 15A is a diagram for explaining a case where processing of a frame in marker extraction processing is stopped, and FIG.
And (C) are diagrams for explaining the method of calculating the approximate center of the marker.

FIGS. 16A and 16B are diagrams for explaining a method of correcting the approximate center when a marker is lost at a frame end, and FIG. 16C is a diagram illustrating another example of such correction. It is a figure showing a table used.

17A is a diagram for explaining a marker for performing a process of correcting the approximate center position in a frame, FIG. 17B is a diagram showing a relationship between a marker and a pattern dot, and FIGS. It is a figure for explaining a marker existence field.

FIG. 18 is a block diagram of a marker center detection unit in FIG. 10;

FIGS. 19A to 19C are diagrams for explaining calculation of a marker position by extrapolation;

FIG. 20A is a diagram for explaining a base marker pair, and FIGS. 20B and 20C are diagrams for explaining a double height.

FIGS. 21A and 21B are diagrams for explaining a method of selecting a reference marker when a base marker pair is selected;
(C) is a diagram for explaining a method of searching for an adjacent marker when a marker candidate is found on only one side of the reference marker.

FIG. 22A is a diagram for explaining the approximate center of a virtual marker, and FIGS. 22B and 22C are diagrams illustrating a method of reselecting a base marker pair when two or more base marker pairs can be selected, respectively. FIG.

FIGS. 23A and 23B are diagrams illustrating a method of reselecting a base marker pair when a code is skewed.

FIGS. 24A and 24B are diagrams for explaining selected marker sets and their storage methods when blocks are not aligned, and FIGS. 24C and 24D are diagrams for explaining the arrangement of blocks; FIG. 11 is a diagram for explaining a selected marker set and a storage method thereof when the setting is performed.

FIGS. 25A and 25B are diagrams for explaining a method of selecting a marker set for selecting a processing target block;

26A is a diagram for explaining a method for reading a block when a reference marker is selected from marker candidates on the scanning direction side, and FIG. 26B is a diagram for explaining a method of reading a block when optical distortion or the like occurs; FIG. 9 is a diagram for explaining a reading order.

FIG. 27A is a diagram for explaining block reselection processing when reading of a selected block has failed;
(B) is a diagram showing the positional relationship between the marker 16 and the pattern dots 14, and (C) is a diagram showing the relationship between the dot reading position and the dot center position.

28A and 28B are diagrams for explaining a method of estimating a data reading reference point using a pattern dot from the approximate center of a marker, and FIG. 28C is a diagram for explaining another estimation method. FIG.

FIGS. 29A and 29B are diagrams for explaining still another data reading reference point estimation method.
(C) and (D) are diagrams for explaining a tree search method.

FIGS. 30A and 30B are diagrams for explaining a limited tree search, and FIG. 30C is a diagram for explaining another arrangement of pattern dots.

31A is a diagram for explaining the operation of the black dot detection unit, FIG. 31B is a diagram for explaining another operation of the black dot detection unit, and FIG. FIG. 6 is a diagram for explaining an operation of a calculation unit.

FIGS. 32A and 32B are diagrams for explaining an operation of estimating a dot center position existence region in a dot center calculation unit, and FIG. 32C is a diagram illustrating reliability of a dot center position existence region and a central representative point. It is a figure for explaining nature.

FIG. 33A is a diagram for explaining calculation of the center of gravity as still another center calculation method in the dot center calculation unit;
(B) to (D) are diagrams for explaining another method of calculating the center of gravity.

34A and 34B are diagrams for explaining a method of calculating a data reading reference point using the least squares method.

35A is a diagram illustrating a coordinate system on a recording medium, and FIG. 35B is a diagram illustrating a coordinate system on an imaging surface.

FIG. 36 is a diagram showing a relationship between a marker pair and ten pattern dots.

FIG. 37 is an operation flowchart in a case where, when there is a dot that is not detected by dot detection, a dot read point is set as the dot center of the dot, and processing is performed assuming that all dots have been detected.

FIGS. 38A and 38B are diagrams showing examples of the arrangement order of blocks and recorded contents of block addresses at that time, and FIG. 38C is a diagram for explaining a data dot reading operation; is there.

[Explanation of symbols]

10: data code, 12: pattern code, 14
... pattern dots, 16 ... markers, 18 ... address dots, 20 ... blocks, 22 ... points, 24 ... dot position reference points, 26 ... guide codes, 28 ... inversion codes, 30 ... correction codes, 32 ... fixed patterns, 3
4 information recording medium, 36 dot code, 38 image input unit, 40 data reading reference point determination unit, 4
2 Data reading unit 44 Data reproducing unit 46
... Imaging unit, 48 ... Equalization and binarization unit, 50 ... Marker extraction unit, 52 ... Marker approximate center calculation unit, 54 ... Marker detection unit, 56 ... Marker center detection unit, 58 ... Marker existence area, 60 ... Frame , 62 ... the center of the marker
64: an area where a marker may be present; 66: center of gravity; 68: circumscribed rectangle; 70: center point of a circumscribed rectangle;
2 ... Frame end, 74: Approximate center of marker detected by deviation due to loss, 76: Correct marker approximate center, 78: Approximate center correction area, 80: Marker partially missing at frame end, 82: Block to be processed, 84 marker pair,
Reference numeral 86: Reference marker selection unit, 88: Marker presence area calculation unit, 90: Marker candidate detection unit, 92: Base marker pair selection unit, 94: Marker pair selection unit, 96: Reference marker, 98: Marker presence area , 100: paired marker, 102: error marker, 104: extrapolated vector (e
xv), 106: Marker search area, 108: Modified extended vector, 110: Base marker pair, 112: Image field of view, 114: Area, 116: Read failure block, 118: Reselection block, 120: Code line, 122 ... Read line, 124: dot read position, 126: dot center position, 128: dot center calculation unit, 130: error minimization unit, 132: black dot detection unit, 134: dot read point, 136: marker approximate center recalculation unit 138: buffer; 140: error detection unit; 142: selection unit; 144: mask;
46: black pixel barycenter in the mask, 150: barycenter representative point position, 152: estimated true center of marker, 154: address dot reading unit, 156: data reading unit, 158: address dot reading unit, 160 ...
Error correction unit, 162: determination unit, 164: address determination unit, 166: last block, 168: data reading point.

Claims (5)

[Claims] An information recording medium in which a code in which a plurality of blocks including a data code corresponding to data of information to be reproduced are arranged is optically readable.
An information reproducing system that optically reads the code to reproduce the information; an image input unit that captures the code; a data reading unit that reads the data from an image captured by the image input unit; Reproducing means for reproducing the information from the data read by the means, wherein the data reading means has an effect of optical distortion from among the plurality of blocks in the image captured by the image input means. An information reproducing system characterized by being configured to select and read a small block. 2. The image processing apparatus according to claim 1, wherein the data reading unit is configured to select and read a block from a center of an image captured by the image input unit.
An information reproduction system according to item 1. 3. The code includes a plurality of blocks having the same data content, and includes a plurality of blocks having the same data content in one frame captured by the image input means. 3. The apparatus according to claim 1, wherein the data reading means is configured to select and read a block having a small influence of optical distortion from the plurality of blocks having the same data content. Information reproduction system. 4. The apparatus according to claim 3, wherein the data code further includes an address dot indicating an address of the block, and the blocks having the same data content have the same address of the block. Information reproduction system. 5. The information reproducing system according to claim 1, wherein the image input unit is configured to continuously capture the code scanned manually.