JP4674853B2 - Image processing component data creation method and image processing component data creation device - Google Patents

Description

  The present invention relates to an image processing component data creation method and an image processing component data creation device for creating image processing component data.

  For example, component data for image processing used in an electronic component mounter needs to be created for each component type to be subjected to image processing. As shown in FIG. 31, the conventional procedure for creating image processing component data is as follows. First, the operator determines the type of component for which the component data is to be created and makes an appropriate decision according to the type of component. An image processing algorithm is selected (procedure 1), and the operator manually inputs data on the shape of the part (body, lead, bump, etc.) according to the type of the part (procedure 2). After that, actually image the part with a camera and perform an image processing test (procedure 3) to check whether the image processing can be performed correctly (procedure 4). If the image processing cannot be performed correctly, Start over from the beginning. As a result, until the image processing can be performed correctly, the operations from the procedure 1 to the procedure 3 are repeated to create component data.

  As described above, in the conventional method for creating component data for image processing, an operator must manually perform operations such as component type discrimination, image processing algorithm selection, component shape data input, and image processing test. (Depending on the result of the image processing test, it is necessary to repeat these operations many times to create one component data). There was a drawback of requiring work time and a lot of knowledge and experience.

  Therefore, as shown in Patent Document 1 (Japanese Patent No. 2941617) and Patent Document 2 (Japanese Patent Laid-Open No. 9-44686), a technique for automating the creation of component data has been proposed.

  Patent Document 1 describes an automatic creation method of component data for determining whether or not a suction posture of a component sucked by a suction nozzle of an electronic component mounting machine is good. In the part data automatic creation method disclosed in Patent Document 1, data such as the dimensions of each part of the part, the number of leads, and the lead interval are calculated as image data by image processing.

Patent Document 2 describes an automatic data creation method for recognizing the position of a component. In the part data automatic creation method of Patent Document 2, the outline of the part image is extracted, the position of the edge end point of each corner portion is obtained for each side of the part, and each side of the basic schematic shape of the part is obtained. And the shape of each corner portion are obtained, and these data are used as the part position recognition data.
Japanese Patent No. 2941617 (3rd page, FIG. 3 etc.) Japanese Patent Laid-Open No. 9-44686 (first page, etc.)

  However, since the methods of Patent Documents 1 and 2 cannot automatically determine the type of component when creating component data, select an appropriate image processing algorithm according to the type of component and The data cannot be created. For this reason, since image data of various types of parts is processed by the same image processing algorithm to generate the part data, there is a drawback in that appropriate part data corresponding to the type of the part cannot be generated.

  The present invention has been made in view of such circumstances. Accordingly, an object of the present invention is to provide an image processing component data creation method and image processing capable of automatically creating appropriate component data according to the type of component. It is to provide a parts data creation device for use.

In order to achieve the above object, the present invention takes in image data of a part for which image processing part data is to be created, and extracts data representing the characteristics of the shape of the whole part from the image data of the part. The type of the part is automatically determined based on the extracted data, and the data including at least one position and size of the body, lead, and bump of the part from the image data of the part according to the automatically determined type of the part The first feature is that image processing component data is generated based on the extracted data, and the data representing the feature of the shape of the entire component is the pixel frequency in the horizontal and vertical directions. The second feature is that data of zero cross pattern frequency is included.

According to the present invention, extracts data from the component image data representative of the shape characteristics throughout the part, since the extracted based on the type of the component data so that automatically detects and automatically discriminated Select an appropriate image processing algorithm according to the type of component, and automatically include data including at least one position and size of the body, lead, and bump of the component from the image data of the component according to the type of component. Therefore, it is possible to automatically create appropriate image processing component data corresponding to the type of component.

In this case, the image data of the shape for the whole part is extracted in advance by taking in the image data of the reference part for each part type, and the data representing the shape characteristic for the whole part for each part type. A stored database is created, and in the part type determination process, data representing the shape characteristics over the entire part is extracted from the image data of the newly captured part, and the extracted data is compared with the data in the database. Thus, it is preferable to automatically discriminate the type of the parts that are the best match. In this way, the accuracy of part type discrimination can be improved.

In the present invention, data representing the characteristics of the shape over the entire part may include data for horizontal direction and the vertical pixel frequency and / or zero-cross pattern frequency is the distinctive feature. Here, the pixel frequency is an appearance frequency of a pixel having a gray value equal to or higher than a predetermined threshold, and the zero-cross pattern frequency is a point that changes from a gray value smaller than the threshold value to a larger gray value, or It means a point where a gray value larger than a threshold value changes to a smaller gray value. If the pixel frequency or the zero cross pattern frequency is used, it is possible to extract data that accurately represents the feature of the shape over the entire part. This makes it possible to automatically determine the type of any component, and to select and create an appropriate image processing algorithm and component data.

Further, in the part type discrimination process, it is preferable to automatically discriminate the part type using a matching method (abbreviation: DP matching) by dynamic programming. DP matching is a typical expansion / contraction matching method, and can calculate the similarity between component data efficiently and accurately.

Hereinafter, an embodiment embodying the best mode for carrying out the present invention will be described.
As shown in FIG. 1, the image processing component data creation device includes a computer 11 such as a personal computer, image data capturing means 12 such as a camera for capturing component image data, and an input device 13 such as a keyboard and a mouse. And a display device 14 such as a liquid crystal display or a CRT, and a storage device 15 for storing programs and data to be described later. The computer 11 executes each program for image processing to be described later, thereby extracting data representing the feature of the shape over the entire part from the image data of the part captured by the image data capturing unit 12, and extracting the extracted data. data including a component type determination means for automatically determining the type of the component on the basis of the image data of the parts of the component according to the type of parts automatically identified body, leads, at least one of position and size of the bumps And functions as a part data creation means for creating part data for image processing based on the extracted data . Hereinafter, processing contents of each program for image processing executed by the computer 11 will be described.

  When the part data creation program of FIG. 2 is started, first, in step 101, the operator waits until the operator operates the input device 13 to instruct the start of part data creation. Thereafter, when the operator instructs the start of component data creation, the process proceeds to step 102, and the image data capturing unit 12 is used to capture the image data of the component for which image processing component data is to be created.

Thereafter, the process proceeds to step 103, where data of a part representing the feature of the shape over the entire part (hereinafter referred to as “feature data”) is extracted from the image data of the captured part. This feature data includes pixel frequency and zero cross pattern frequency data in the horizontal and vertical directions, which will be described later. After that, the extracted feature data is compared with the feature data in the reference part data database stored in the storage device 15 to automatically determine the type of the parts that are the best match. This makes it possible to automate the process of selecting an image processing algorithm according to the type of component. The discrimination of the component type is performed by using a matching method (abbreviation: DP matching) by dynamic programming described later. The reference component data database is created by a database creation program shown in FIG.

  Thereafter, the process proceeds to step 104, where image processing component data (data such as the position and size of the body, leads, and bumps) of the component is extracted from the image data of the component in accordance with the determined type of component.

  Thereafter, the process proceeds to step 105, and the extracted image processing component data is stored in the image processing component data database stored in the storage device 15 in association with the type of the component.

  The database creation program in FIG. 3 is a program for creating a database of reference component data. When this program is started, first, in step 201, image data of reference parts is captured using the image data capturing means 12.

  After that, the process proceeds to step 202, where the feature data extraction program is executed to extract the feature data of the part (body, lead, bump, etc.) representing the feature of the part from the image data of the reference part by the method described later. To do. At this time, the pixel frequency in the horizontal direction and the vertical direction and the zero cross pattern frequency, which are data suitable for representing the feature of the shape over the entire part, are calculated as the feature data.

  Thereafter, the process proceeds to step 203, and the extracted feature data of the reference part (horizontal and vertical pixel frequency and zero cross pattern frequency pattern) is stored in the reference part data database stored in the storage device 15. Save it in association with the part type.

  4 is a program for executing the processing of steps 102 to 103 of the component data creation program of FIG. When this program is started, first, in step 301, the image data capturing unit 12 is used to capture the image data of the component for which image processing component data is to be created.

  Thereafter, the process proceeds to step 302, and the feature data extraction program is executed in the same manner as in step 202 of the database creation program in FIG. Extract feature data of the part to be represented (body, lead, bump, etc.). At this time, the pixel frequency and zero cross pattern frequency in the horizontal direction and the vertical direction are calculated as feature data.

  Thereafter, the process proceeds to step 303, where the feature data of the component for which image processing component data is to be created is compared with the feature data in the reference component data database stored in the storage device 15, and The type is automatically detected. At this time, the type of component is determined using a matching method (abbreviation: DP matching) by dynamic programming described later.

  The pixel frequency calculation program of FIG. 5 is a program executed in step 103 of FIG. 2, step 202 of FIG. 3, and step 302 of FIG. 4, and calculates the pixel frequency as follows. When this program is started, first, in step 401, image data of a part is read. Thereafter, the process proceeds to step 402, where a block of a part portion (part existence range) is cut out from the image of the entire screen of the display device 14. With regard to cutting out the part block, the operator may operate the input device 13 by looking at the part image displayed on the screen of the display device 14 and cut out the part block. A partial block cutout program may be executed to automatically cut out a part block.

  In the component part block cutout program of FIG. 7, first, in step 501, a histogram of the entire image is obtained, and a threshold value is calculated by a discriminant analysis method. Thereafter, the process proceeds to step 502, the image is scanned in the horizontal and vertical directions, the frequency of appearance of pixels having a gray value equal to or greater than the threshold value is counted, and a pixel frequency graph in the horizontal and vertical directions is created.

  Thereafter, the process proceeds to step 503, the pixel frequency graph in the horizontal direction and the vertical direction is smoothed by a Gaussian filter, and then the process proceeds to step 504, in which the pixel frequency is equal to or higher than a predetermined frequency in the pixel frequency graph in the horizontal direction and the vertical direction. Is determined as a part block (part existence range) and cut out.

  After cutting out the part part block, the process proceeds to step 403 in FIG. 5 to calculate the rotation angle of the part part block with respect to the reference coordinates (XY coordinates), and the horizontal and vertical directions of the part part block are the horizontal direction of the reference coordinates. Rotation correction is performed so as to match the (X direction) and the vertical direction (Y direction). Regarding this rotation correction, the operator may look at the image of the component displayed on the screen of the display device 14 to determine the rotation angle of the component, and operate the input device 13 to correct the rotation angle of the component. The rotation angle of the component part block may be automatically corrected by a method described later.

  After correcting the rotation angle of the block of the part part, the process proceeds to step 404, and a threshold value for binarization is calculated using an appropriate binarization method such as binarization of Otsu, and this threshold value is calculated. Is used to binarize the image of the part block (see FIG. 6).

  Thereafter, the process proceeds to step 405, where the image is scanned in the horizontal and vertical directions, and the frequency (pixel frequency) at which a pixel having a gray value equal to or greater than the threshold exists on each horizontal scanning line is counted. The pixel frequency in the horizontal direction is defined as Pdx, the frequency (pixel frequency) at which pixels having a gray value equal to or greater than the threshold value are present on each vertical scanning line is counted, and the pixel frequency in the vertical direction is defined as Pdy ( (See FIG. 6).

  The zero cross pattern frequency calculation program of FIG. 8 is a program executed in step 103 of FIG. 2, step 202 of FIG. 3, and step 302 of FIG. 4, and calculates the zero cross pattern frequency as follows. When this program is started, first, in step 601, image data of a part is read. Thereafter, the process proceeds to step 602, where a block of a part portion (part existence range) is cut out from the image of the entire screen of the display device 14 by the same method as step 402 of the pixel frequency calculation program of FIG.

  Thereafter, the process proceeds to step 603, where the rotation angle of the part block relative to the reference coordinates (XY coordinates) is calculated in the same manner as in step 403 of the pixel frequency calculation program of FIG. Rotation correction is performed so that the vertical direction matches the horizontal direction (X direction) and the vertical direction (Y direction) of the reference coordinates.

  Thereafter, the process proceeds to step 604, where a threshold value for binarization is calculated using an appropriate binarization method such as binarization of Otsu, and the block of the part portion is calculated using this threshold value. Binarize the image.

  Thereafter, the process proceeds to step 605 to detect a zero cross edge. Here, the zero-cross edge means a point where the gray value smaller than the threshold value changes to a larger gray value, or a point where the gray value larger than the threshold value changes to a smaller gray value.

  Thereafter, the process proceeds to step 606, where the frequency (zero cross pattern frequency) at which zero cross edges exist on each horizontal scanning line is counted, and this horizontal zero cross pattern frequency is set to Pzx, and on each vertical scanning line. The frequency at which zero-cross edges exist (zero-cross pattern frequency) is counted, and the zero-cross pattern frequency in the vertical direction is set as Pzy (see FIG. 9).

Next, a specific example of the rotation angle detection method used in step 403 of the pixel frequency calculation program of FIG. 5 and step 603 of the zero cross pattern frequency calculation program of FIG. 8 will be described. Hereinafter, two rotation angle detection methods will be described. Needless to say, other rotation angle detection methods may be used.
[Rotation angle detection method (part 1)]
In general, it is often better to determine the component angle based on the external shape of the component.

  In consideration of this point, the rotation angle detection method (part 1) extracts the outline portion of the region having a gray value equal to or higher than the threshold value in the captured image of the component by the procedure shown in FIGS. This is a method of detecting the rotation angle by judging the linearity.

  Hereinafter, a specific example of this rotation angle detection method (part 1) will be described with reference to FIGS. The rectangle in each figure represents the lead of the IC.

First, as shown in FIGS. 10 and 11, the search is started from the edge of the image, and a position where a pixel having a predetermined threshold value or more appears is obtained (examples of searching from left to right in FIGS. 10 and 11). Is shown).
Thereafter, as shown in FIG. 12, the position of the outer portion is used to perform Hough transform to detect a straight line.

  The above processing is performed in each of the four directions of the image, as shown in FIGS. Then, the one having the highest linearity is selected from the results retrieved from the four directions, and the angle of the straight line is set as the component angle.

The features of the rotation angle detection method (part 1) described above are as follows.
(1) Generally, when line segment detection is performed using the Hough transform, calculation time is proportional to the number of calculation points (for example, the number of edge points of the target image).
(2) Usually, in a method for obtaining the angle of an object in an image, an edge point in the image is used, or a point binarized with an appropriate threshold value is used.
(3) In many cases, the rotation angle of mounted components can be detected with relatively high accuracy even if only the points constituting the outermost shape are used.

  Among these features, the rotation angle detection method (part 1) extracts the points constituting the outermost contour and uses only those points for the calculation of the Hough transform, while reducing the calculation cost, Can be obtained with high accuracy.

  In the specific example described with reference to FIGS. 10 to 16, the angle detection is performed using the points searched from the image ends in the four directions. This means that the search must be performed from the four directions. is not. You may employ | adopt the method of employ | adopting the highest linearity among the straight lines calculated | required by searching from 2 directions or 3 directions.

  Here, I would like to emphasize the feature that the Hough transform usually requires a lot of calculation time, but the angle of the mounted component etc. can be obtained accurately even using only the points constituting the outermost shape. This means that the number of points used for the Hough transform is reduced and high-speed and high-precision angle detection is performed.

  In addition, the object is not only expressed in one lump (parts imaged with a backlight), but also expressed in multiple lumps (IC lead parts or BGA parts imaged with a front light) The rotation angle detection method (part 1) also functions effectively.

[Rotation angle detection method (2)]
The rotation angle detection method (part 2) uses a boundary line of a region having a gray value equal to or greater than a threshold value in a captured image of a part, and a circumscribed rectangle (hereinafter referred to as “nearest neighbor rectangle”) that includes the boundary line most. ). Here, “included in a rectangle” means existing on a line segment of the rectangle.

  FIG. 17 shows a simple example. For the minimum circumscribed rectangle often used for detecting the rotation angle of an object, it is considered that only the longest line segment in the region and other corner portions are included in the circumscribed rectangle. On the other hand, for the circumscribed rectangle having the angle to be recognized, it is considered that all line segments other than the longest line segment are included in the circumscribed rectangle. In this case, the boundary line included in the circumscribed rectangle is the largest.

In this rotation angle detection method (part 2), it is necessary to determine whether one point with a boundary line is included in the circumscribed rectangle.
However, usually, the boundary line of the target area is not a straight line but has undulations.
Therefore, it is not possible to obtain a correct result only by determining whether or not the line is completely on the circumscribed rectangular line segment.

  Accordingly, as shown in FIG. 18, a method is used in which it is determined that a boundary line existing within a certain distance from each line segment of the circumscribed rectangle is included in the circumscribed rectangle.

  At this time, all the points on the boundary line existing in the area determined to be included in the circumscribed rectangle may be evaluated equally, or an evaluation according to the distance from the circumscribed rectangle to the point on the boundary line is given. May be.

  For example, if an evaluation function as shown in FIG. 19 is given, the evaluation is performed using an evaluation value proportional to the distance between the circumscribed rectangle and the boundary point. An evaluation function such as a Gaussian function is also conceivable, but any arbitrary evaluation function can be used.

Next, a specific example of the rotation angle detection method (part 2) will be described with reference to FIGS. The rectangle in each figure represents the lead of the IC.
First, as shown in FIG. 20, after binarizing an image using an appropriate threshold value, a boundary line of the region is obtained. In addition, you may obtain | require a boundary line using another method.

Thereafter, the following processing is executed while changing the search angle from −10 degrees to 10 degrees in increments of 0.1 degrees.
First, as shown in FIG. 21, when the search angle is θ, the boundary line group is rotated by −θ.

Thereafter, as shown in FIG. 22, a circumscribed rectangle in which each side is parallel to the horizontal and vertical axes is created. At this time, a circumscribed rectangle may be created while changing the angle of the rectangle.
Thereafter, as shown in FIG. 23, out of the points of the boundary tracking result, the points existing within a certain distance from each side of the circumscribed rectangle are extracted, and the evaluation calculation is performed. This evaluation calculation is performed by defining an evaluation function that gives a high score when the distance from each side of the circumscribed rectangle is short and gives a low score when it is far as shown in FIG.

In the experiment conducted by the present inventor, an evaluation function was used in which a point within 5 pixels from each side of the circumscribed rectangle is effective, and 5 points are obtained at a distance of 0 pixels and 0 points are obtained at a distance of 5 pixels.
In this case, the score of the search angle θ is a total score of evaluation values calculated at all points existing below a certain distance from each side of the circumscribed rectangle.

  When the processing described above is repeated while changing the search angle θ and the evaluation calculation is performed for all the search angles, the search angle having the highest evaluation score is obtained, and this angle is set as the rotation angle of the component. Thereby, the correct angle can be detected regardless of the shape of the component region.

  Next, in step 103 of the part data creation program of FIG. 2 and step 303 of the part type discrimination program of FIG. 4, the type of the part is selected using a matching method (abbreviation: DP matching) by dynamic programming. A specific example of automatic discrimination will be described.

  In this DP matching, a reference is selected from a reference component data database, and unknown components are classified with a small number of reference patterns. Here, the following five methods 1 to 5 can be considered as a method of selecting a reference.

[Method 1]
In order to perform an evaluation equivalent to selecting the maximum number when viewed up to a certain distance d ′, the equation of FIG. 25 is used.

[Method 2]
The one having the smallest total distance from the smallest DP distance to n is selected by the equation of FIG.

[Method 3]
The one with the smallest average DP distance is selected by the equation of FIG.

[Method 4]
As shown in FIG. 28, the one near the center (for example, methods 1, 2 and 3) is selected by applying a Parzen window to obtain a continuous frequency distribution when there are few samples. Thereby, it is expected that the classification result is stabilized even when the samples are discrete.

[Method 5]
As shown in FIG. 29, one-dimensional clustering is performed to select one close to the center of each cluster (for example, methods 1, 2, and 3). This is expected to stabilize the classification result.

Next, a determination method by DP matching will be described.
Using Pdx and Pdy obtained from the pixel frequency and Pzx and Pzy obtained from the zero-cross pattern frequency, classification is performed according to the distance at which DP matching of each pattern is performed. A conceptual diagram of DP matching is shown in FIG. By performing DP matching, it is possible to measure the difference when the positional deviation between the two vectors (graphs) is minimized. In the case of complete coincidence, the distance D (I, J) becomes 0, and the distance D (I, J) becomes farther (larger) as it is different.

  At this time, the classification that minimizes the distance D (I, J) by DP matching is determined as the classification of the component. In this case, it was found that when Tmax and Rmax are normalized to 1 in FIG. 30, the classification discrimination accuracy increases.

  According to the present embodiment described above, since the type of the part is determined from the image data of the part, an appropriate image processing algorithm corresponding to the determined type of the part is selected, and the type of the part is selected. Accordingly, data representing the characteristics of the part can be automatically extracted from the image data of the part, and appropriate part data corresponding to the type of the part can be automatically created.

In this embodiment, the data representing the feature of the shape of the entire part includes both the horizontal and vertical pixel frequencies and the zero-cross pattern frequency, but only one of them is included. also not good.

1 is a block diagram illustrating a configuration of an image processing component data creation device according to an embodiment of the present invention. It is a flowchart which shows the flow of a process of a components data creation program. It is a flowchart which shows the flow of a process of a database creation program. It is a flowchart which shows the flow of a process of a component kind discrimination | determination program. It is a flowchart which shows the flow of a process of a pixel frequency calculation program. It is a figure explaining the result of a pixel frequency calculation process. It is a flowchart which shows the flow of a process of a component part block cutout program. It is a flowchart which shows the flow of a process of a zero cross pattern frequency calculation program. It is a figure explaining the result of a zero crossing pattern frequency calculation process. It is a figure explaining the external shape detection process from the left direction in the rotation angle detection method (the 1). It is a figure explaining the external shape detection result from the left direction in the rotation angle detection method (the 1). It is a figure explaining the straight line detection process in a rotation angle detection method (the 1). It is a figure explaining the external shape detection processing method from the left direction in the rotation angle detection method (the 1). It is a figure explaining the external shape detection processing method from the right direction in the rotation angle detection method (the 1). It is a figure explaining the external shape detection processing method from the upper direction in the rotation angle detection method (the 1). It is a figure explaining the external shape detection processing method from the downward direction in the rotation angle detection method (the 1). It is a figure explaining the creation method of the minimum circumscribed rectangle and the nearest rectangle in the rotation angle detection method (the 2). It is a figure explaining the area | region determined to be contained in the circumscribed rectangle in the rotation angle detection method (the 2). It is a figure explaining the evaluation function used with the rotation angle detection method (the 2). It is a figure explaining how to obtain the boundary line in the rotation angle detection method (part 2). It is a figure explaining the process which rotates the boundary line group in the rotation angle detection method (the 2). It is a figure explaining the method of creating the circumscribed square in the rotation angle detection method (the 2). It is a figure explaining the method to extract the point which exists below a certain fixed distance from each edge | side of the circumscribed rectangle in a rotation angle detection method (the 2). It is a figure explaining the specific example of the evaluation function used with the rotation angle detection method (the 2). It is a figure explaining the reference selection method 1 of DP matching. It is a figure explaining the reference selection method 2 of DP matching. It is a figure explaining the reference selection method 3 of DP matching. It is a figure explaining the reference selection method 4 of DP matching. It is a figure explaining the reference selection method 5 of DP matching. It is a figure which illustrates notionally the component type discrimination | determination method by DP matching. It is a flowchart which shows the flow of the conventional component data creation work.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 11 ... Computer (part kind discrimination means, part data creation means), 12 ... Image data taking means, 13 ... Input device, 14 ... Display device, 15 ... Storage device

Claims (6)

Processing to capture image data of a component to be created for image processing component data;
Extracting data representing the characteristics of the shape of the entire part from the image data of the part, and a part type determination process for automatically determining the type of the part based on the extracted data;
Data including at least one position and size of the body, lead, and bump of the part is extracted from the image data of the part according to the type of the part that has been automatically determined, and image processing part data is extracted based on the extracted data. viewing including the process of creating,
The image processing component data creation method, wherein the data representing the feature of the shape of the entire component includes pixel frequency and / or zero cross pattern frequency data in the horizontal and vertical directions . A database in which image data of reference parts is extracted in advance for each part type, data representing shape characteristics over the entire parts is extracted, and data representing the shape characteristics over the entire parts is stored for each part type. Create
In the component type determination process, data representing the shape characteristics over the entire component is extracted from the image data of the newly captured component, and the extracted data is compared with the data in the database to find the component with the best match. 2. The image processing component data creation method according to claim 1, wherein the type is automatically determined. 3. The image processing component data creation method according to claim 1, wherein in the component type determination process, a component type is automatically determined using a matching method based on dynamic programming. Image data capturing means for capturing image data of a component to be created for image processing component data;
Extracting data representing the characteristics of the shape over the entire part from the image data of the part, a part type determining means for automatically determining the type of the part based on the extracted data;
Data including at least one position and size of the body, lead, and bump of the part is extracted from the image data of the part according to the type of the part that has been automatically determined, and image processing part data is extracted based on the extracted data. only it contains the component data creation means for creating,
The image processing component data creation apparatus characterized in that the data representing the feature of the shape of the entire component includes pixel frequency and / or zero cross pattern frequency data in the horizontal and vertical directions . A database that stores data representing the shape characteristics of the entire part for each part type is provided, and the part type determination unit is configured to determine the shape of the entire part from the image data of the part newly captured by the image data capturing unit. 5. The image processing component data generation apparatus according to claim 4 , wherein data representing features is extracted, and the extracted data is compared with the data in the database to automatically determine the type of the component that best matches the data. . 6. The image processing component data creation apparatus according to claim 4, wherein the component type discrimination means automatically discriminates the type of the component using a matching method based on dynamic programming.

Images