JPH07325878A - Recognizing method for character - Google Patents

Abstract

PURPOSE:To provide a recognizing method for a character capable of accurately recognizing the character even when dust and stains are present at the time of reading the character by a machine such as an OCR or the like. CONSTITUTION:Patterns 2 for recognition formed by arranging square black areas with an equal interval in the direction of the side for which the size of the equal interval is changed corresponding to the character are written with the characters 1. Then, at the time of reading them, by calculating the interval between the black areas of the read patterns for the recognition, what kind of the character 1 the pattern 2 for the recognition is corresponding to is recognized and the character 1 is recognized. A white area and the black area can be replaced. Also, when the average value of the ones calculated for plural times is adopted as the interval between the black areas of the read patterns 2 for the recognition, the probability of erroneous recognition due to the attachment of the dust and the stains is made further smaller.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition method for recognizing a character without error when the character is read by a machine such as an OCR (optical character reader).

[0002]

2. Description of the Related Art Conventionally, a direct method and an indirect method are known as methods for reading written characters by a machine. In the direct method, the correspondence between a character and the characteristics of the character itself is defined in advance, the characteristics are extracted from the image of the entire read character, and then which character is the characteristic in light of the correspondence. Is a method of examining and recognizing characters.

In the direct method, since the image of the entire character is used as information for recognition, the amount of information is large, and therefore the number of characters to be recognized can be increased. But,
This direct method has the following drawbacks. The first drawback is that each character must retain the characteristics of that character, which greatly limits the shape of the character. The second drawback is the variation in print density,
If there is dust or dirt, the characteristic part of the character cannot be read correctly and the character is erroneously recognized. The third drawback is that since a large amount of information is used for recognition, a high-performance device capable of high-speed processing is required for quick recognition.

The indirect method was conceived in an attempt to improve the above drawbacks even a little. In the indirect method, a special character for the indirect method is defined in advance, and what is written using the special character is read and recognized. The special character referred to here is a character in which a specific code determined for recognizing the character is attached in the middle of a line forming the normal character or in the vicinity of the normal character. As the unique code, for example, a plurality of black circles are used. Depending on the letters, the relative arrangement of these black circles is unique. When a character is read, what kind of character the character is is recognized by checking the relative arrangement of the black circles.

According to the indirect method, the character is not recognized from the image information of the entire character but the character is recognized based on the contained unique code, so that the amount of information processed for the recognition is the direct method. Less than Also, there are less restrictions on the shape of characters. Note that, as the conventional literatures regarding the indirect method, for example, Japanese Patent Publication No. 38628/1993 and Japanese Patent Publication No. 3-6
There is a publication such as 3115.

[0006]

However, in the above-mentioned conventional indirect method, since the character is recognized based on the unique code included in or near the character, one of the unique codes is recognized. However, there is a problem that the character cannot be accurately recognized when it is not read due to variations in print density, dust or dirt attached.
An object of the present invention is to solve the above problems.

[0007]

In order to solve the above-mentioned problems, in the character recognition method of the present invention, regular tetragonal attention areas are formed by arranging reversal areas at equal intervals in the direction of their sides. And by additionally writing a recognition pattern in which the size of the equal interval is changed according to the character to the character, reading the recognition pattern, and calculating the interval between the attention areas of the read recognition pattern, We decided to recognize what kind of characters the characters are.

If the average value of a plurality of calculated areas is used as the interval between the attention areas of the read recognition pattern, erroneous recognition due to adhesion of dust or dirt can be further prevented.

[0009]

[Operation] A recognition pattern in which regular tetragonal regions of interest are formed at equal intervals in the direction of the side thereof with an inversion region between them and the size of the regular intervals is changed according to the character Write along with the letters. Then, when they are read, the distance between the attention areas of the read recognition pattern is calculated to know what kind of character the recognition pattern corresponds to, and the character is recognized. The attention area may be a black area or a white area. If the black area is adopted, the inversion area is a white area. If the white area is adopted, the inversion area is used. Is a black area.

[0010]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a diagram showing characters recognized in the present invention. Reference numeral 1 is a character, and 2 is a recognition pattern. In the present invention, when the character 1 is displayed, a predetermined recognition pattern 2 is additionally written to distinguish it from other characters. Therefore, the types of the recognition pattern 2 are determined in advance by the number of characters to be recognized. For example, the recognition target is “1”, “2”, “3”, “4” as shown in FIG.
If there are only four numbers, the four recognition patterns are predetermined. If the number is from "0" to "9", 10 recognition patterns are predetermined.

FIG. 2 is a diagram showing an example (a) of a basic pattern used in the present invention and a recognition pattern (b) composed of the basic pattern. 3 is a black area and 4 is a white area. This basic pattern is a pattern in which black areas 3 of regular tetragons (rectangular or square; square is used in FIG. 2) of the same size are arranged at equal intervals in the direction of each side. is there. The black area 3 is the attention area. As will be described in detail later, the key serving as a key for character recognition is the interval between the attention areas. Figure 2
The recognition pattern 2 in (b) is created by successively repeating the basic pattern. The pattern created in this way is used as the recognition pattern 2 in FIG.

FIG. 4 is a diagram showing an example of another basic pattern used in the present invention. After all, the black area is adopted as the attention area. 4A is a basic pattern in which square black areas 3 of the same size are arranged in the direction of each side at intervals of a length twice that of the side, and FIG. 4B shows the same. This is a basic pattern in which black areas 3 having a size of square are arranged in the direction of each side at intervals of a length three times that of the side.

FIG. 3 is a diagram showing a hardware configuration for carrying out the method of the present invention. 5 is an image reading device, 6 is a bus, and 7 is a bus.
Is an image output device, 8 is a CPU (central processing unit), 9
Is memory. The image reading device 5 reads the character 1 to be recognized together with the recognition pattern 2. The read signal is converted into a digital image signal and stored in the memory 9 through the bus 6. The CPU 8 performs a character recognition operation (see FIGS. 5 and 6) based on the digital image signal stored in the memory 9. The recognized character is output from the image output device 7 as needed.

Next, the recognition operation of the present invention will be described. FIG. 5 is a diagram for explaining the recognition operation in the present invention, and FIG. 6 is a flowchart showing the procedure of the recognition operation in the present invention. As shown in FIG. 5, it is assumed that the recognition pattern that is also written on the read character is the same pattern as that in FIG. That is, it is assumed that the black region 3 is a pattern arranged in the direction of each side at an interval equal to the length of the side. Hereinafter, a description will be given according to the flowchart of FIG.

Step 1 ... A parameter N indicating the number of times the process has proceeded to step 13 of this flowchart (the number of times the distance has been calculated) is set, and the value is set to zero. Step 2: An arbitrary point A is set in the read recognition pattern. The coordinates of the point A are generated by the CPU 8.
FIG. 5 shows a case where the arbitrarily generated position of the point A takes various positions in the recognition pattern. The points (A ₁ etc.) of the letters with the subscripts attached to A are specific examples of the points A. The same applies to the other characters B to D.

The position of the point A can be roughly divided into two. In the first case, there is a black area 3 with a completely white area 4 on the left (points A ₁ and A ₂ correspond thereto), and in the second case, other cases. (Point A ₃ , point A ₄ , and point A ₅ correspond thereto). In addition, the case of the point A _{3 is} a case where the black region 3 does not exist no matter how far it goes to the left. In the case of the points A ₄ and A ₅ , the black area 3 exists on the left side, but the black area 3 that separates the completely white area 4 does not exist.

Step 3 ... Starting from point A, the coordinates are advanced in the X direction (to the left). Step 4 ... Check for black dots. Point A is A
If you take a position like ₃ , there is no black dot no matter how far you go. In that case, the process returns to step 2 and another arbitrary point A is newly set and redone. Step 5 ... If there is a black dot, mark that point as B, and then X
Advances the coordinates in the direction. In the case of the point A ₁ taken in the black area 3, the adjacent dot is a black dot, and this is B ₁ . In the case of the point A ₂ taken in the white area 4, the black dot appears when it reaches the adjacent black area 3, and that point becomes B ₂ .

Step 6 ... Further advance the coordinates to check whether there is a white dot. When advancing from the black dot, the black region 3 always exits and enters the white region 4, so that the white dot always exists. Step 7 ... The point of the white dot is C (C ₁ , C _2, etc.)
And further advance the coordinates in the X direction. From here,
If a black dot is found after that, a certain black area 3
A complete white area 4 lying between one and the next black area 3
(Distance between black areas) is crossed. By calculating the crossing distance (interval), it is possible to recognize which character the pattern for recognition corresponds to, and thus it is possible to recognize which character.

Step 8: Check for black dots.
If there is, proceed to step 13 and set that point to D (D ₁ ,
D ₂ etc.). Step 9 ... If not, the process returns to the previous point B, and the coordinates are advanced from there in the Y direction. For example, when the point A is taken at the position of A ₄ or A ₅ , the point C becomes C ₄ or C ₅ , but there is no black dot in the X direction beyond this. Therefore, in these cases, the process returns to the point B ₄ or B ₅ and proceeds in the Y direction.

Step 10: Check for white dots.
Point B is a black dot, so if you proceed in the Y direction, you will eventually reach a white dot. Step 11: When the white dot is reached, the point is set to E (E ₄ , E _5, etc.) and the coordinates are further advanced in the Y direction. Step 12 ... Check for black dots. If not, the process returns to step 2, another arbitrary point A is taken, and the process is repeated. If there is, go to step 13 and set that point as D (D ₄ , D _5, etc.).

Step 13: The distance between the point D and the point C or the point E, which is the white starting point before proceeding to the point D, is obtained. If the point D is obtained after proceeding in the X direction, the distance between the points C and D is obtained, and if the point is bent in the Y direction and the point D is obtained, the distance between the points E and D is obtained. This allows
The distance between the black areas 3 of the read recognition pattern 2 is obtained. Step 14 ... Since the distance (interval) between the black regions can be calculated once starting from a certain point A, the value of the parameter N is increased by one.

Step 15 ... The value of N is a predetermined N ₀
Find out if it is equal to. If not equal, step 2
Return to. Steps related to parameter N (1, 14, 1
5, 16) are not indispensable and are provided in order to minimize the probability of erroneous recognition.

That is, since the intervals between the black regions, which are clues for character recognition in the present invention, appear everywhere in the recognition pattern, most of them appear even if they are partially hidden by dust or dirt. ing. Since the point A for calculating the interval is located at an arbitrary position, the probability of being located at a position such as dust is small. Therefore, in most cases, the characters are recognized correctly.

However, if dust or dirt adheres to the vicinity of the point A, they are treated as if they were black areas, and therefore the intervals between black areas cannot be calculated accurately. However, another point A that is far away is taken one after another, and a plurality of times (N ₀ times)
If the calculation is made, most of them will be points free of dust and dirt, so in most cases the distance will be accurate. Therefore, if the average value is taken for recognition, the probability of misidentification is further reduced.

Step 16 ... N obtained as described above
Take the average of ₀ distances. Step 17 ... Since the distances between the black areas of the respective recognition patterns in use are known in advance, they are compared with the average value obtained this time, and the read recognition pattern corresponds to which character. By determining whether there is a character, what kind of character the read character is.

In the embodiment, the black area is adopted as the attention area. However, as described above, a pattern in which the attention area is the white area and the black areas are arranged at equal intervals is used. It can also be a recognition pattern.

[0027]

As described above, according to the character recognition method of the present invention, square black areas are formed at equal intervals in the direction of the side, and the size of the equal intervals is set according to the character. The different recognition patterns are written together for the character, and when they are read, the character is recognized by calculating the interval between the black areas of the read recognition pattern.

The cue for character recognition is the space between the black areas, but since it exists everywhere in the recognition pattern, there are variations in the print density of the recognition pattern, and there are some differences. Even if dust or dirt is attached to the, the clue of character recognition is not lost. Since it is rare that the points taken for calculating the intervals coincide with the places where dust or dirt adheres, accurate character recognition can be performed in most cases. The greater the number of intervals calculated, the more accurate the recognition.

[Brief description of drawings]

FIG. 1 is a diagram showing characters recognized by the present invention.

FIG. 2 is a diagram showing an example of a basic pattern used in the present invention and a recognition pattern composed of the basic pattern.

FIG. 3 is a diagram showing a hardware configuration for carrying out the method of the present invention.

FIG. 4 is a diagram showing an example of another basic pattern used in the present invention.

FIG. 5 is a diagram for explaining a recognition operation according to the present invention.

FIG. 6 is a flowchart showing a procedure of a recognition operation of the present invention.

[Explanation of symbols]

1 ... Characters, 2 ... Recognition pattern, 3 ... Black area, 4 ... White area, 5 ... Image reading device, 6 ... Bus, 7 ... Image output device,
8 ... CPU, 9 ... Memory

Claims (2)

[Claims] 1. A recognition pattern in which regular tetragonal regions of interest are formed at equal intervals in the direction of the sides thereof with an inversion region therebetween, and the size of the regular intervals is varied according to the character. A method for recognizing a character, which is characterized in that the character is recognized by reading the recognition pattern together with the character and calculating an interval between the attention areas of the read recognition pattern. 2. The character recognition method according to claim 1, wherein an average value of a plurality of calculated recognition patterns is adopted as an interval between the attention areas of the read recognition pattern.