JP2009250707A - Master image selection method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To realize a selection of a master image which can be used commonly for an acceptance decision of multiple kinds of assembly components by a pattern matching and which can perform an acceptance decision with high reliability. <P>SOLUTION: A correlation value of each photographed image among a group of photographed images (assembly components) for one kind is calculated, and then, an average value X of the correlation values, a variance value σx, and a lowermost value Xmin of the correlation values (=X-3*σx) are calculated. Similarly for other kinds of the component group, an average value Y of the correlation values and a variance value σy are calculated, and a uppermost value Ymax (=Y-3*σy) is calculated. Then a difference (=Xmin-Ymax) of the lowermost value Xmin and the uppermost value Ymax is calculated, respectively, and an image wherein the difference becomes maximum is chosen as a master image. An acceptance decision with high reliability wherein a variation in assembly components is considered is realized by introducing an idea of 3σ in calculating the lowermost value Xmin and the uppermost value Ymax. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

The present invention relates to a master image selection method for selecting a master image used as a reference image when determining whether or not there is an abnormal assembly of an inspection object by pattern matching.
The presence / absence of an abnormality in the inspected object indicates, for example, whether or not an assembly part (product type) with respect to the assembly position is appropriate, or whether or not the assembly position or state of the part is appropriate.

  Conventionally, as this type of technique, an optimal reference pattern / reference image (master image) is selected by repeating calculation for obtaining a degree of coincidence between captured images (see Patent Document 1), or a plurality of reference images are selected. There has been a technique for selecting a pattern image by performing an overlay process (see Patent Document 2).

JP-A-9-251536 JP 2006-329679 A

However, all of the above prior arts are effective in selecting an optimal master image in inspection between parts of a single type (IC leads, etc.), but selection of an optimal master image in inspection between multiple types of parts. I couldn't.
For example, in a multi-mixed line represented by an automobile assembly line, it is necessary to inspect various types of parts with the same image inspection device. The image could not be selected.

  The present invention has been made in view of the above circumstances, and can be used in common for determining whether or not there is an assembly abnormality of a plurality of types of inspection objects. In addition, the inspection object within a specific type is highly accurate. It is an object to provide a master image selection method capable of selecting a master image that can be judged as normal (non-defective product) and that can reliably determine abnormality (incorrect product) with respect to an inspection product of a product other than one type. And

The above-mentioned problem is solved by adopting a master image selection method having the following aspects.
As with the claims, each aspect is divided into sections, each section is numbered, and is described in a form that cites the numbers of other sections as necessary. This is merely for the purpose of facilitating the understanding of the present invention, and the technical features described in this specification and combinations thereof should not be construed as being limited to those described in the following sections. . In addition, when a plurality of items are described in one section, it is not always necessary to employ the plurality of items together, and it is also possible to take out only a part of the items and employ them.

Of the following items, (1) corresponds to claim 1 and (2) corresponds to claim 2.
The item (3) is not the claimed invention.

(1) In the master image selection method for selecting a master image to be used as a reference image when determining whether there is an abnormality in assembly of the inspection object by pattern matching of a photographed image of the inspection object attached thereto, Pattern matching is performed using a specific captured image in a specific captured image group in a plurality of types of captured image groups each having a plurality of captured images of the inspection object as a temporary master image to calculate a correlation value with other captured images The average value, the variance value, and the minimum value of the correlation value are calculated, and the correlation value with each captured image in each captured image group other than the specific captured image group is patterned using the temporary master image. By executing matching, a correlation value with each captured image is calculated, and an average value, a variance value, and a maximum upper limit value of the correlation value are calculated, and the temporary master image And sequentially changing the captured images in the specific captured image group to perform both pattern matching to calculate the lower limit value and the upper limit value of the correlation value, and the correlation value when using each temporary master image A master image selection method comprising: calculating a difference between the lowest and upper limit values of each of the two and determining the temporary master image used when executing the pattern matching that maximizes the difference as a master image.
“Assembly abnormality of inspection object” refers to an abnormality in the type of inspection object, for example, the assembled part is different from a predetermined part, and the mounting position of the inspection object Or an abnormal mounting state, for example, a part is mounted at a position different from a predetermined position or is mounted in a loose state.
The master image selected (determined) according to the invention of this section has a strong correlation with all of the captured images in one captured image group (specific captured image group), and the one captured image. Outside of the group, the image is in the one imaged image group that is weakly correlated with all the imaged images in the same group and suitable for imaged image determination (existence / non-existence determination of inspected object assembly). In short, the master image selected according to the invention of this section indicates whether there is an abnormal assembly of the inspection object by pattern matching with all the captured images in the plurality of captured image groups, and the captured images in one captured image group. This is one photographed image in one photographed image group that is optimal for use and determination.
(2) In the master image selection method for selecting a master image to be used as a reference image when determining the presence / absence of assembly abnormality of an inspection object by pattern matching of a photographed image of the inspection object attached thereto, For all the captured images in a specific captured image group in a plurality of types of captured image groups each having a plurality of captured images of the inspection object, pattern matching is performed using the specific captured image in all the captured images as a temporary master image, A first step of calculating a correlation value with the provisional master image, a second step of calculating an average value and a variance value of the entire correlation value obtained by the first step, and each of those other than the specific photographed image group Perform pattern matching with the temporary master image for all the captured images in the captured image group, except for the specific captured image group A third step of calculating a correlation value between each captured image in each captured image group and the temporary master image, and a fourth step of calculating an average value and a variance value of all the correlation values obtained by the third step; The first step to the fourth step are repeatedly executed while sequentially replacing the temporary master image with a captured image other than the specific image in the specific captured image group, and each captured image in each captured image group A fifth step of calculating a correlation value with each of the temporary master images, a sixth step of calculating an average value and a variance value of the entire correlation values obtained by the fifth step, the second step, the fourth step From the average value and the variance value calculated in the step and the sixth step, respectively, the lowest limit value of the correlation value for determining the presence or absence of assembly abnormality in the specific captured image group, the specific captured image group, and Other shooting Temporary master images used to calculate the maximum correlation value for determining the presence or absence of an assembly error between image groups and to calculate the difference between them, and to perform the pattern matching that maximizes the difference And a seventh step of determining as a master image. A master image selection method comprising:
The invention of this section embodies the invention of section (1).
The “inspection abnormality of the inspection object” indicates an abnormality in the type of the inspection object or an abnormality in the attachment position or attachment state of the inspection object, as in the invention of the item (1).
(3) The lowest limit value of the correlation value is calculated by X-3 * σx when the average value of the correlation values is X and the variance value is σx, and the maximum upper limit value of the correlation values is the correlation value The master image selection method according to item (1) or (2), wherein Y is calculated by Y−3 * σy when the average value of Y is Y and the variance is σy.
“*” Represents multiplication (the same applies hereinafter).
According to the invention of this section, when determining pass / fail of assembled parts by pattern matching, variation (3σ) of the parts is taken into consideration, and a non-defective product can be determined with high accuracy for a specific one type of part, A master image that can reliably determine a wrong product can be selected for a component of a type other than the one type, and the reliability can be enhanced by using it for non-defective product or wrong product determination (inspection). Further, the time required for selecting such a master image is short.

  According to the present invention, it can be used in common for determining whether or not there is an abnormality in assembling of plural types of inspection objects (parts) by pattern matching, and normal (non-defective) with respect to a specific type of inspection object. ) And a master image that can reliably determine an abnormality (incorrect product) can be selected for an inspected product other than the one type.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. In addition, the same code | symbol shows the same or an equivalent part between each figure.
FIG. 1 is a block diagram showing an embodiment of an apparatus to which a master image selection method according to the present invention is applied.
In this figure, reference numeral 11 denotes a digital camera (hereinafter abbreviated as a camera) composed of a color or monochrome CCD (Charge Coupled Device) camera or the like for photographing the object to be inspected 10, and the photographed image is output as image data.
A personal computer (hereinafter abbreviated as a personal computer) 12 has a function of capturing a captured image (image data) from the camera 11 and storing it in the storage device 13, a pattern matching processing function for executing pattern matching on the captured image, And a master image selection processing function for selecting an optimum master image. Here, the pattern matching is a method for detecting by comparing both images whether there is an image similar to or close to a master image (reference image) prepared in advance in a captured image of the object 10 to be inspected. .

In the present embodiment, the personal computer 12 also has a function of correcting the positional deviation of the captured image due to variations in the shooting direction and the like.
The pattern matching processing function has a function of specifying a pattern to be matched and setting the processing area.
The master image selection processing function has at least the following functions.
(1) Even if there are variations in the inspection object for each product type (group) and variations in the shooting environment (conditions) of the inspection object, all of the captured images are within the allowable range in advance. To import as a processing target for
(2) Function to sort captured images by type (group)
(3) Function to register one specified photographed image as a temporary master image and execute pattern matching processing not only on the same kind but also on other kinds of photographed images to calculate the correlation value
(4) Function to calculate and compare the average value and variance value of the calculated correlation values
(5) From the calculated average value and variance of the correlation values, it is possible to determine whether a product is good for the same product type (pass / fail judgment for the product type), and to determine a wrong product (product type error) for other product types. Function for Selecting a Master Image The above functions operate by executing a program stored in the storage device 13 of the personal computer 12.
Note that the captured image, temporary master image, master image, and the like captured by the personal computer 12 can be displayed on the display 14 in a timely manner.

Next, the operation of the above-described apparatus will be described with reference to FIGS.
This apparatus is an apparatus that selects a master image that is used as a reference image when determining whether there is an abnormality in assembly of the inspection object 10 by pattern matching of a captured image of the assembled inspection object 10.
Here, the presence / absence of an assembly abnormality of the inspection object 10 indicates, for example, whether or not an assembly part (product type) is appropriate for the assembly position, or whether or not the assembly position of the part is appropriate. In the present embodiment, it indicates whether or not the type of assembly part with respect to the assembly position is acceptable, that is, whether the part assembled at the assembly position is a correct part (no assembly abnormality) or an incorrect part (with assembly abnormality).
In addition, this apparatus determines whether or not there is an abnormal assembly for each of the inspected objects 10 in a plurality of m types of captured image groups having a plurality of n photographed images of the inspected object 10. This is a device for selecting a master image to be used.
In the illustrated example, there are n photographed images of the object 10 to be inspected, each of which includes three photographed image groups of tires A to C with different manufacturers (see FIG. 2). The device is a master image selection device that is used to determine whether or not a correct type of tire among these three types of tires is attached to the vehicle.

As shown in FIG. 3, first, in step 101, a first captured image A1 of a captured image group (hereinafter simply referred to as a group) A is set as a temporary master image (temporary master image).
In step 102, the first processing group (processing target group) is set to A.
In step 103, the temporary master image A1 is used to perform pattern matching on each of the captured images A1 to An in the processing target group A to calculate a correlation value, and the average value (X) of all the correlation values is calculated. A dispersion value (σx) is calculated.

In step 104, the lowest limit value Xmin = X−3 * σx of the entire correlation value (correlation value distribution) is calculated. This minimum lower limit value Xmin is an estimated value of a lower limit correlation value at which non-defective product determination can be made for each of the captured images A1 to An in the own group A.
In step 105, pattern values are executed for the captured images B1 to Bn and C1 to Cn of the groups B and C that are not to be processed to calculate correlation values, and the average value (Y) and variance values of the entire correlation values are calculated. (Σy) is calculated. This is to see the relationship between the captured images B1 to Bn and C1 to Cn of the other groups B and C.

In step 106, the maximum upper limit value Ymax = Y−3 * σy of the entire correlation value (correlation value distribution) is calculated. This maximum upper limit value Ymax is an estimated value of an upper limit correlation value that can be determined as a wrong product for each of the captured images B1 to Bn and C1 to Cn in the other groups B and C.
An example of the values calculated in steps 103 to 106 is shown in FIG.
In step 107, the difference (Z = Xmin−Ymax) between the lowest limit value and the highest limit value of the correlation value when the temporary master image (captured image A1) is used is calculated. The greater the difference Z between Xmin and Ymax calculated in step 107, the higher the accuracy of non-defective product / wrong product determination, so that an optimal master image can be obtained. Therefore, if the newly calculated difference Z is larger than the maximum difference value Zmax so far, the maximum value Zmax is updated to the newly calculated difference Z (if Z> Zmax, Z is updated to Zmax).
In step 108, the update is performed, and the temporary master image at this time is stored in the storage device 13 (see FIG. 1) as an optimal master image candidate.

In step 109, it is determined whether all the captured images A1 to An in the processing target group A have been set as temporary master images. If the determination result is YES, the process proceeds to step 110, and if the determination result is NO, step 111 is executed and the process proceeds to step 103. That is, steps 103 to 109 are repeated until the nth (final) captured image An is set as the temporary master image, and Z and the temporary master image (optimum master image candidate) are updated.
In step 110, the temporary master image stored in the storage device 13 (see FIG. 1) at this time is held as the optimum master image selected from the captured images A1 to An in the processing target group A.

In step 112, it is determined whether all groups A to C have been processed. If the determination result is YES, the entire process is terminated, and if NO, step 113 is executed and the process proceeds to step 103. That is, the above steps 103 to 112 are repeated until the nth (final) captured image Cn of the group C (final processing target group) is set as the temporary master image. Update master image candidate).
It is to be noted that group B and group C are processed after group A. In the groups B and C, the first set as the temporary master image is the taken images B1 and C1, and the last set is the taken images Bn and Cn.
The temporary master images (three temporary master images) stored and held in the storage device 13 (see FIG. 1) at the final time of all the processes are the captured images A1 to An, B1 to the processing target groups A to C. It is an optimal master image used for non-defective product / wrong product selection for each of the processing target groups A to C selected from Bn and C1 to Cn.

In the present embodiment described above, the correlation value of each captured image in a group of one type of captured image (inspection object 10) is calculated, the average value X of the correlation values, the variance value σx, and the lowest limit of the correlation values The value Xmin (= X−3 * σx) is calculated, the average value Y and the variance value σy of the correlation values are calculated in the same manner for other types of component groups, and the maximum upper limit value Ymax (= Y−3 *). σy) is calculated. Then, a difference (= Xmin−Ymax) between the lower limit value Xmin and the upper limit value Ymax is obtained, and the image when the difference becomes maximum is used as the master image.
According to this, it is possible to select a master image that can be used in common for determining whether or not there is an abnormality in assembly of plural types of inspection objects 10 by pattern matching (e.g. whether assembly parts are acceptable). In addition, the variation (3σ) of the inspected object 10 is taken into consideration in the above determination, the non-defective product can be determined with high accuracy for the inspected object 10 in one specific type, and inspected for products other than the one type For the object 10, it is possible to select a master image that can reliably determine an incorrect product. Therefore, if the master image according to the present embodiment is used for non-defective product / wrong product determination, the reliability of the non-defective product / wrong product determination can be improved.
In addition, since the master image in this embodiment can be easily selected by calculation using the personal computer 12, there is an advantage that the time required for selecting the master image can be shortened.

It is a block diagram which shows one Embodiment of the apparatus with which the master image selection method by this invention was applied. It is a figure which illustrates the list of the picked-up image of the to-be-inspected object memorize | stored according to the group in the memory | storage device of the apparatus same as the above in tabular form. It is a flowchart which shows the selection process of the optimal master image in an apparatus same as the above. It is a figure which shows an example of the value calculated in the process same as the above in a table format.

Explanation of symbols

10: Inspected object, 11: Digital camera, 12: Personal computer (PC), 13: Storage device.

Claims (2)

In the master image selection method for selecting the master image used as the reference image when determining the presence or absence of the assembly abnormality of the inspected object by pattern matching of the captured image of the assembled inspected object,
A pattern matching is performed using a specific captured image in a specific captured image group in a plurality of types of captured image groups each having a plurality of captured images of the inspection object as a temporary master image, and a correlation value with another captured image is obtained. Calculating the lowest value of the average value, the variance value, and the correlation value, and using the temporary master image as a correlation value with each captured image in each captured image group other than the specific captured image group The pattern matching is performed to calculate the correlation value with each captured image, the average value, the variance value, and the maximum upper limit value of the correlation value are calculated,
The captured images in the specific captured image group used as the temporary master image are sequentially changed to execute the pattern matching to calculate the minimum and maximum correlation values, respectively, and use each temporary master image. A difference between the lowest and highest correlation values is calculated, and the master image used to perform the pattern matching that maximizes the difference is determined as a master image. Image selection method. In the master image selection method for selecting the master image used as the reference image when determining the presence or absence of the assembly abnormality of the inspected object by pattern matching of the captured image of the assembled inspected object,
Pattern matching is performed on all the captured images in a specific captured image group in a plurality of types of captured image groups each having a plurality of captured images of the inspection object, using the specific captured image in all the captured images as a temporary master image. A first step of calculating a correlation value with the temporary master image;
A second step of calculating an average value and a variance value of the entire correlation values obtained in the first step;
For all the captured images in each captured image group other than the specific captured image group, pattern matching with the temporary master image is performed, and each captured image in each captured image group other than the specific captured image group A third step of calculating a correlation value with the temporary master image;
A fourth step of calculating an average value and a variance value of all the correlation values obtained in the third step;
The first to fourth steps are repeatedly executed while sequentially replacing the temporary master image with a captured image other than the specific image in the specific captured image group, and each captured image in each captured image group A fifth step of calculating a correlation value with each of the temporary master images;
A sixth step of calculating an average value and a variance value of the entire correlation value obtained in the fifth step;
From the average value and the variance value calculated by the second step, the fourth step and the sixth step, respectively, the lowest limit value of the correlation value for determining the presence or absence of assembly abnormality in the specific photographed image group, The maximum value of the correlation value for determining the presence or absence of assembly abnormality between the specific photographed image group and the other photographed image groups is calculated and the difference between them is calculated, and this difference is maximized. A seventh step of determining the temporary master image used at the time of executing the pattern matching as a master image;
A master image selection method comprising:

Images