JP2018105756A - Master image data creation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a master image data creation device with a new structure which can create more desirable master image data, for example.SOLUTION: The master image data creation device according to an embodiment includes: a deformed image data creation unit for creating deformed image data, which is obtained by deforming second master image data in relation to first master image data, by fitting processing on the basis of detection of feature points for the first and second master image data; and a correction unit for correcting the first master image data on the basis of the deformed image data.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a master image data creation device.

  Conventionally, a technique for creating master image data from a plurality of image data in image inspection is known.

JP 2005-142552 A

  In this type of technology, for example, it is meaningful if a master image data creation device having a novel configuration capable of creating master image data with less inconvenience can be obtained.

  In the master image data creation device of the embodiment, the second master image data and the second master image data are associated with the first master image data by fitting processing based on feature point detection for the second master image data and the second master image data. A deformed image data creating unit that creates deformed image data obtained by deforming the master image data, and a correcting unit that corrects the first master image data based on the deformed image data.

FIG. 1 is a schematic and exemplary block diagram illustrating a schematic configuration of a master image data creation device according to an embodiment. FIG. 2 is an exemplary flowchart illustrating a processing procedure performed by the master image data creation device according to the embodiment. FIG. 3 is a schematic and exemplary conceptual diagram illustrating an offset of base data in preprocessing by the master image data creation device of the embodiment. FIG. 4 is a schematic and exemplary conceptual diagram showing an offset different from that of FIG. 3 of the base data in the preprocessing by the master image data creation device of the embodiment. FIG. 5 is a schematic and exemplarily conceptual diagram illustrating an area that is a target of similarity calculation by the master image data creation device of the embodiment. FIG. 6 is a table showing a setting example of determination criteria according to the combination of the master data or update master data class and the matching data class by the master image data creation device of the embodiment.

  Hereinafter, exemplary embodiments of the present invention are disclosed. The configuration and control of the embodiment shown below, and the operation and result (effect) brought about by the configuration and control are examples. The present invention can be realized by other than the configurations and controls disclosed in the following embodiments. Further, the present invention can obtain various effects including derivative effects obtained by basic configuration and control.

  FIG. 1 is a schematic configuration diagram of a master image data creation device 1. The master image data creation device 1 includes an arithmetic processing unit 10 and a storage unit 20.

  The arithmetic processing unit 10 includes a data reading unit 11, a preprocessing unit 12, a conforming data creation unit 13, a similarity calculation unit 14, an adoption determination unit 15, a master data update unit 16, an arithmetic processing management unit 17, and the like. Yes. The arithmetic processing unit 10 is, for example, a CPU (central processing unit) or a controller, and executes the arithmetic processing in accordance with software (program), whereby the data reading unit 11, the preprocessing unit 12, and the conforming data creation unit 13, the similarity calculation unit 14, the adoption determination unit 15, the master data update unit 16, the arithmetic processing management unit 17, and the like. At least a part of the arithmetic processing unit 10 is hardware such as a DSP (digital signal processor), an ASIC (application specific integrated circuit), an FPGA (field programmable gate array), or a PLD (programmable logic device). Also good.

  The storage unit 20 stores master data MD, base data BD, update master data RMD, conformance data FD, employment status data AD, and the like. The storage unit 20 may include, for example, a nonvolatile rewritable storage device such as a hard disk drive (HDD) or a volatile rewritable storage device such as a random access memory (RAM).

  The master data MD, the base data BD, the update master data RMD, the conformance data FD, and the adoption state data AD all correspond to a rectangular imaging area or a calculation area in the imaging area. For example, in the i-axis direction (imax) × (jmax) including a predetermined number (imax) of pixels i = 1 to imax and including a predetermined number (jmax) of j = 1 to jmax in the j-axis direction orthogonal to the i-axis. Data of values (digital data) in each pixel arranged two-dimensionally. Usually, the master data MD, the base data BD, the update master data RMD, and the matching data FD are narrower than the imaging area.

  The master data MD is, for example, image data obtained by photographing a non-defective product to be inspected. The base data BD is, for example, image data obtained by photographing a non-defective product to be inspected. In the present embodiment, the arithmetic processing unit 10 corrects the master data MD serving as a reference using the base data BD. The update master data RMD is data in which the master data MD is corrected, and is data that is updated as needed in the arithmetic processing. The matching data FD is data obtained by transforming the base data BD to match the master data MD based on the feature point detection. The value (luminance value, intensity) of each pixel of the matching data FD is calculated by complementation from the value of each pixel of the base data BD. The adoption state data AD is data (for example, a flag) indicating whether or not the matching data FD is used for correction in each pixel. Note that the master data MD is an example of the base data BD in that it is the source of the final update master data RMD (final correction data of the master data MD), and the base data BD is the master data BD. It can be said that it is an example of data MD. The master data MD or the updated master data RMD is an example of first master image data. The base data BD is an example of second master image data. The matching data FD is an example of deformed image data.

  FIG. 2 is a flowchart of calculation processing by the calculation processing unit. The arithmetic processing unit 10 corrects the master data MD according to the procedure shown in FIG. The arithmetic processing unit 10 corrects the master data MD using a plurality of base data BD, and creates updated master data RMD. The updated master data RMD when a predetermined condition is satisfied is used for image inspection as corrected master data, that is, reference data. That is, an image inspection apparatus (not shown) detects the presence / absence of an abnormality by comparing photographed data with reference data. In the following, for example, a calculation process in a case where an abnormal region that is white on the inspection target surface is generated, such as a tire, has a black surface as an inspection target surface. The abnormality indicates, for example, dirt that appears at unspecified places, printing, marks, and the like.

  In FIG. 2, the arithmetic processing unit 10 functions as the data reading unit 11 and reads the master data MD (S1). Next, the arithmetic processing unit 10 functions as the data reading unit 11 and reads the base data BD (S2). The data reading unit 11 reads different base data BD every time S2 is executed.

  Next, the arithmetic processing unit 10 functions as the preprocessing unit 12 and executes preprocessing for the master data MD or the updated master data RMD and the base data BD (S3). The preprocessing in S3 can include processes such as masking, offset (shift), and feature point detection, for example. The master data MD is used in the first process of S3 to S8, and the updated master data RMD is used in the second and subsequent processes.

  Masking includes masking of areas that are not subject to detection, masking of areas that are not subject to arithmetic processing related to correction of master data MD, and the like.

  The offset (shift) is alignment by shifting the base data BD relative to the master data MD or the update master data RMD. 3 and 4 exemplify the offset of the base data BD with respect to the master data MD or the update master data RMD by the preprocessing unit 12. As shown in FIGS. 3 and 4, the direction in which the base data BD is shifted may be the i direction or the j direction. When the inspection target surface is an annular surface or the like, the preprocessing unit 12 divides the base data BD into two regions BD1 and BD2 as shown in FIG. Can be moved to the opposite side along the circumferential direction (j direction) of the annular surface, that is, the opposite side of the region BD1 where the deviation is small. Note that the preprocessing unit 12 may extract feature points of the master data MD or the update master data RMD and the base data BD by image processing, and may perform alignment based on the feature points. Further, the preprocessing unit 12 may enlarge or reduce the base data BD according to the master data MD or the update master data RMD. In addition, the preprocessing unit 12 complements the value of each pixel of the base data BD aligned corresponding to the master data MD or the updated master data RMD from the value of each pixel of the base data BD before alignment. Depending on the situation, it may be calculated.

  For the feature point detection, for example, a known image processing technique such as corner detection by FAST feature detection or the like is used. Further, edge detection by a Sobel operator or the like may be performed in order to make it easier to detect corner points.

  Next, the arithmetic processing unit 10 functions as the matching data creation unit 13 and creates matching data FD corresponding to the master data MD or the update master data RMD based on the base data BD (S4). In S4, the matching data creation unit 13 first detects the feature points of the base data BD corresponding to the feature points of the master data MD or the updated master data RMD. Next, the matching data creation unit 13 associates the feature points with each other by a known fitting process (pattern matching process). For the association between feature points, for example, a known method is used such that a vector amount is calculated by an optical flow, and the nearest vector amount is selected by using the degree of similarity between a peripheral direction component and a norm as an index. Here, the feature points of the base data BD are associated with the feature points of the master data MD or the update master data RMD so that the image of the base data BD is as close as possible to the image of the master data MD or the update master data RMD. It is equivalent to deforming (distorting). The value of each pixel in the matching data FD obtained by deformation of the base data BD is calculated by interpolation processing of the value of each pixel of the base data BD. As a method for deforming and interpolating an image, for example, a known method such as projective transformation is used. The matching data creation unit 13 is an example of a deformed image data creation unit.

絶対差分総和Ｒ_ＳＡＤが０に近いほど類似度が高く、１に近いほど類似度が低い。
（２）正規化相互相関Ｒ_ＺＮＣＣ（ＺＮＣＣ：zero-mean normalized cross-correlation）

ここで、Ｍ_Ａは、マスタデータＭＤまたは更新マスタデータＲＭＤの画素の値の領域Ａｍ内の平均値、Ｆ_Ａは、適合データＦＤの画素の値の領域Ａｆ内の平均値である。正規化相互相関Ｒ_ＺＮＣＣが１に近いほど類似度が高く、０に近いほど類似度が低い。平均値で正規化される分、全体的な明るさや暗さが異なる場合、すなわち一方が他方よりもオフセットしているような場合であっても、画像の明暗の変化の類似度、すなわち画像の類似度が反映されやすい。
なお、正規化相互相関Ｒ_ＺＮＣＣに替えて、次の式（３）で示される正規化相互相関Ｒ_ＮＣＣ（ＮＣＣ：normalized cross-correlation）を用いてもよい。

正規化相互相関Ｒ_ＮＣＣが１に近いほど類似度が高く、０に近いほど類似度が低い。
なお、代表画素は、領域Ａｍ，Ａｆの中心でなくてもよい。 Next, the arithmetic processing unit 10 functions as the similarity calculation unit 14 and calculates the similarity between the master data MD or the update master data RMD and the matching data FD (S5). In S <b> 5, the similarity calculation unit 14 calculates the following two similarities (1) and (2) for each pixel (representative pixel) in the calculation target region. Here, as an example, the representative pixel of the master data MD or the updated master data RMD is M (i, j), the representative pixel of the matching data FD is F (i, j), and the similarity calculator 14 calculates the similarity. FIG. 5 shows a pixel value in an 11 × 11 area Am centered on M (i, j) and a pixel in an 11 × 11 area Af centered on F (i, j). Based on the value of.
(1) The absolute difference between the sum total _{R SAD (SAD: sum of absolute} differences)

The closer the absolute difference sum R _SAD is to 0, the higher the similarity is, and the closer the value is to 1, the lower the similarity is.
(2) normalized cross-correlation _{R ZNCC (ZNCC: zero-mean} normalized cross-correlation)

Here, M _A is the average value of the area Am of the values of the pixels of the master data MD or updated master data RMD, F _A is the average value of the area Af of pixel values of matching data FD. The closer the normalized cross-correlation _RZNCC is to 1, the higher the similarity, and the closer to 0, the lower the similarity. Even if the overall brightness and darkness are different by the average value, that is, when one is offset from the other, the similarity of the change in the lightness of the image, that is, the image Similarity is easily reflected.
Instead of the normalized cross-correlation _{R ZNCC,} normalized cross-correlation _R NCC represented by the following formula (3) (NCC: normalized cross -correlation) may be used.

The closer the normalized cross-correlation R _NCC is to 1, the higher the similarity is, and the closer it is to 0, the lower the similarity is.
Note that the representative pixel may not be the center of the areas Am and Af.

Next, the arithmetic processing unit 10 functions as the adoption determining unit 15 and corrects the master data MD or the updated master data RMD by the matching data FD for each pixel of the matching data FD based on the similarity calculated in S5. Whether to use each pixel of the matching data FD for correction of the master data MD or the updated master data RMD (S6). In this determination, the adoption determination unit 15 classifies the pixels of the master data MD or the update master data RMD and the pixels of the matching data FD according to the pixel values, for example, as follows. Note that the threshold values used for classification are the same between the master data MD or the update master data RMD and the matching data FD.
(Class 1) Pixel whose pixel value is equal to or less than the threshold value Thl As described above, in the present embodiment, the calculation process is performed when the inspection target surface is based on black and dark at a normal position, that is, when the luminance value is small. Is illustrated. Therefore, it can be said that a pixel classified into class 1 is highly likely to be included in a normal region.
(Class 2) Pixels whose pixel value is greater than the threshold value Thl and less than or equal to the threshold value Thh Pixels that are not included in class 1 or class 3 are classified as class 2.
(Class 3) Pixel whose pixel value is larger than the threshold Thh A pixel classified as class 3 is highly likely to be included in an abnormal area.

  Then, as illustrated in FIG. 6, the adoption determining unit 15 performs, for each pixel, according to cases A to C of combinations of the pixel class of the master data MD or the update master data RMD and the pixel class of the matching data FD. It is determined whether or not to adopt based on criteria of different similarity.

(Case A) When master data MD or update master data RMD is class 3 and conforming data FD is class 1 In case A, there is a high possibility that the pixels of master data MD or update master data RMD are abnormal. At the same time, there is a high possibility that the pixels of the matching data FD are normal. Therefore, in case A, the threshold value is set so that correction by the pixels of the matching data FD is easier than in other cases. For example, the following formulas (4-1) and (4-2) for each of the two similarities are used.
R _SAD ≧ Thsh (4-1)
R _ZNCC ≦ Thzl (4-2)
Here, Thsh is a threshold value of the absolute difference sum R _SAD , and is 0.7 as an example. Thzl is a threshold value of the normalized cross-correlation _RZNCC , and is 0.3 as an example. In case A, for example, when either of formula (4-1) or formula (4-2) is satisfied (OR condition), the adoption determining unit 15 determines the pixel (representative pixel) of the conforming data FD. Is determined to be used for correction, and the value of the pixel in the employment status data AD is set to “1”. On the other hand, when neither of the expressions (4-1) and (4-2) are satisfied, the adoption determining unit 15 determines that the pixel of the conforming data FD is not used for correction, and adopting state data The value of the pixel in AD is set to “0”.

(Case B) When master data MD or update master data RMD is class 1 and conforming data FD is class 3 In case B, there is a high possibility that the pixels of master data MD or update master data RMD are normal. At the same time, there is a high possibility that the pixels of the matching data FD are abnormal. Therefore, in case B, the threshold value is set so that correction by the pixels of the matching data FD is less likely to be performed than in other cases. For example, the following formulas (5-1) and (5-2) for each of the two similarities are used.
R _SAD ≧ Thsl (5-1)
R _ZNCC ≦ Thzh (5-2)
Here, Thsl is a threshold value of the absolute difference sum R _SAD and is smaller than Thsh in case A (Thsl <Thsh), and is 0.1 as an example. Thzh is a threshold value of the normalized cross-correlation _RZNCC , which is larger than Thzl in case A (Thzh> Thzl), and is 0.9 as an example. In case B, for example, when the expressions (5-1) and (5-2) are satisfied (AND condition), the adoption determining unit 15 corrects the pixel (representative pixel) of the matching data FD. The value of the pixel in the employment status data AD is set to “1”. On the other hand, if either of the expressions (5-1) and (5-2) is not satisfied, the adoption determining unit 15 determines that the pixel of the conforming data FD is not used for correction, and the adoption state. The value of the pixel in the data AD is set to “0”.

(Case C) When neither Case A nor Case B In Case C, an intermediate threshold (normal threshold) between Case A and Case B is set. For example, the following formulas (6-1) and (6-2) for each of the two similarities are used.
R _SAD ≧ Thsm (6-1)
R _ZNCC ≦ Thzm (6-2)
Here, Thsm is a threshold value of the absolute difference sum R _SAD , which is smaller than Thsh in case A and larger than Thsl in case B (Thsh>Thsm> Thsl), and is 0.3 as an example. Thzm is a threshold value of the normalized cross-correlation _RZNCC , which is larger than Thzl in case A and smaller than Thzh in case B (Thzl <Thzm <Thzh), and is 0.7 as an example. In Case C, for example, when either of Expression (6-1) or Expression (6-2) is satisfied (OR condition), the adoption determination unit 15 determines the pixel (representative pixel) of the conforming data FD. Is determined to be used for correction, and the value of the pixel in the employment status data AD is set to “1”. On the other hand, when neither of the expressions (6-1) and (6-2) is satisfied, the adoption determining unit 15 determines that the pixel of the conforming data FD is not used for correction, and the adoption state data The value of the pixel in AD is set to “0”. The adoption determination unit 15 is an example of a determination unit. It can be said that the plurality of determination criteria for case A to case C shown in FIG. 6 are a plurality of determination criteria that differ depending on the pixel value of the master data MD or the update master data RMD. It can be said that there are a plurality of judgment criteria different depending on the value of the pixel of the FD, and a plurality of different criteria depending on the combination of the pixel value of the master data MD or the update master data RMD and the pixel value of the matching data FD It can be said that it is a criterion.

Next, the arithmetic processing unit 10 functions as the master data update unit 16 and updates the master data MD or the update master data RMD (S7). In S7, the master data updating unit 16 uses the following formula (7-1) for the master data MD or the pixel determined to be used for the correction of the adaptation data FD according to the result of the adoption determination in S6. The pixel value of the update master data RMD is corrected (updated).
V _R = (V _M + V _F ) / n (7-1)
Here, n is the number of times of overlap (for example, n = 2).
That is, the master data update unit 16 updates the pixel value of the update master data RMD to an average value of the pixel value of the master data MD or the update master data RMD and the pixel value of the matching data FD.

However, the master data update unit 16, in the case A, with respect to the conformance data FD based on the first base data BD, the pixel of the master data MD or the update master data RMD according to the following equation (7-2) Correct (update) the value of.
V _R = V _F (7-2)
In case A, there is a high possibility that the pixels of the master data MD or the updated master data RMD are included in the abnormal area and the pixels of the matching data FD are included in the normal area. Therefore, it is possible to eliminate the influence of the pixel data in the abnormal region on the subsequent arithmetic processing by the correction (update) using the equation (7-2). In this case, since there is no overlap, it is not necessary to increase the number of times n in the pixel at the corresponding location. The master data update unit 16 is an example of a correction unit.

Next, the arithmetic processing unit 10 functions as the arithmetic processing management unit 17 and determines whether or not to execute arithmetic processing based on another matching data FD (S8). In more detail, in S8, the arithmetic processing management unit 17 first calculates the adoption rates R _a. The adoption rate _Ra is the ratio of the number of pixels determined to be used for correction in each of the cases A to C in S6 with respect to the total number of pixels that can be used for correction in the matching data FD. For example, when the following condition 1 (formula (8-1)) is satisfied, the arithmetic processing management unit 17 can determine that the process based on the other matching data FD is not executed (No in S8).
(Condition 1) R _a ≧ Thr (8-1)
Here, Thr is a threshold of the adoption rate Ra, and is 0.98 as an example. If No in S8, the process proceeds to S9.

  On the other hand, in the case of Yes in S8, that is, when the condition 1 is not satisfied, the process returns to S3, and the arithmetic processing unit 10 executes the arithmetic processes of S3 to S7 again. However, in S4, the matching data creation unit 13 uses the same parameters as the base data BD that is the same as the base data BD that created the matching data FD used in the previous calculation process, and uses different parameters from those used before. The matching data FD different from that used in the above is created. The arithmetic processing unit 10 executes S5 to S7 based on the different matching data FD. In this case, the matching data creation unit 13 uses, for example, pseudo random numbers or the like, which are different from the previous ones (parameters related to the fitting process (deformation), for example, the search range width, high The matching data FD is generated based on the detection threshold of the feature points, etc., and the same matching data FD is not created from one base data BD in a plurality of loops of S3 to S8 corresponding to one base data BD. To.

  Further, in a plurality of loops of S3 to S8 corresponding to one base data BD, the correction using the pixel value of the matching data FD is executed only once for each pixel. That is, the arithmetic processing unit 10 is a pixel that has already been corrected by the pixel value of any of the matching data FD corresponding to one base data BD in the loop of S3 to S8 corresponding to one base data BD, that is, For pixels for which the employment status data AD has been set to “1” in the past, correction based on the pixel value of the matching data FD is not executed. Thereby, for example, S5 to S7 can be omitted for a pixel whose adoption state data AD value is “1”, so that the calculation amount in the calculation processing unit 10 can be reduced and the calculation load can be reduced. .

In S8, the arithmetic processing management unit 17 does not execute processing based on another matching data FD corresponding to one base data BD when the following condition 2 (formula (8-2)) is satisfied. (No in S8).
(Condition 2) R _r ≦ Thl (8-2)
Here, R _r is _a change rate of the adoption rate R _a in a plurality of loops of S3 to S8 corresponding to one base data BD, for example, R _a (n) −R _a (n−1) Yes (n: number of loops) Thl is a threshold value corresponding to R _r and is, for example, 0.01. The condition 2 may be any condition as long as it can detect that the adoption rate Ra is in an invariable state (saturated state), that is, saturated, in a plurality of loops of S3 to S8 corresponding to one base data BD. It is not limited to (8-2). If No in S8, the process proceeds to S9.

  Note that the loop of S3 to S8 may be executed as follows. That is, the arithmetic processing unit 10 creates a plurality of different matching data FD corresponding to one base data BD, and calculates the adoption rate Ra for each matching data FD. And the arithmetic processing part 10 performs correction | amendment (update) based on the conformity data FD in order with the high adoption rate Ra. In other words, in the arithmetic processing corresponding to one base data BD, the master data updating unit 16 applies the one base data BD in descending order of the adoption rate Ra under the condition that the already corrected pixel value is not corrected. The above-described correction (update) based on a plurality of corresponding different matching data FD can be executed. Thereby, the correction of the master data MD by the base data BD may be executed more efficiently or more quickly.

  Next, the arithmetic processing unit 10 functions as the arithmetic processing management unit 17 and determines whether or not to execute arithmetic processing based on another base data BD (S9). In S9, the arithmetic processing management unit 17 prepares a plurality of base data BDs used for correcting the master data MD (or the updated master data RMD), for example, when all the base data BDs are not used. Return to S2. This is an example in the case of Yes in S9. The arithmetic processing management unit 17 may determine whether to execute processing based on another base data BD based on a different condition.

  As described above, in the present embodiment, the matching data creation unit 13 (deformed image data creation unit) performs master data MD or update master data RMD (first master image data) and base data BD (second data). The fitting data FD (deformed image data) obtained by deforming the base data BD in association with the master data MD or the updated master data RMD is generated by fitting processing based on feature point detection on the master image data). The master data update unit 16 (correction unit) corrects the master data MD or the update master data RMD based on the matching data FD. Therefore, according to the present embodiment, the master data MD or the updated master data RMD can be corrected by the matching data FD that is aligned with higher accuracy. Therefore, as master data (reference data) used for image inspection by comparison with sample images, for example, update with higher accuracy or higher quality than master data corrected by data with insufficient alignment Master data RMD can be obtained.

  In the present embodiment, the adoption determination unit 15 (determination unit) performs correction by the master data update unit 16 for each pixel based on a comparison between the master data MD or the update master data RMD and the matching data FD. Determine whether or not. Therefore, according to the present embodiment, for example, it is not necessary to perform the correction when it is not suitable for the correction, so that the update master data RMD with higher accuracy or higher quality can be obtained.

  In the present embodiment, the adoption determination unit 15 (determination unit) determines whether to perform correction for each pixel based on the similarity between the regions Am and Af. Therefore, according to the present embodiment, for example, when the degree of similarity is low and the fitting is insufficient, correction is not performed, so that updated master data RMD with higher accuracy or higher quality can be obtained. .

  In the present embodiment, the adoption determination unit 15 (determination unit) uses a determination criterion set according to the combination of the pixel value of the master data MD or the update master data RMD and the pixel value of the matching data FD. Based on this, it is determined whether or not correction is to be performed for each pixel. Therefore, according to the present embodiment, for example, when there is a high possibility that the pixels of the matching data FD are included in the abnormal region, it is possible to set a state in which correction is difficult to be performed. When there is a high possibility that the pixels of the update master data RMD are included in the abnormal area, it is possible to set a state in which correction is easily performed. Therefore, the update master data RMD with higher accuracy or higher quality can be set. Can be obtained.

  In the present embodiment, the master data updating unit 16 can correct the master data MD or the updated master data RMD for each pixel by using a plurality of matching data FD created from one base data BD. . Therefore, according to the present embodiment, for example, update master data RMD with higher accuracy or higher quality can be obtained corresponding to one base data BD.

  As mentioned above, although embodiment and the modification of this invention were illustrated, the said embodiment is an example and is not intending limiting the range of invention. The above embodiment can be implemented in various other forms, and various omissions, replacements, combinations, and changes can be made without departing from the spirit of the invention. The above embodiments are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof. In addition, the configurations and shapes of the plurality of embodiments may be partially exchanged. In addition, the specifications (direction, shape, size, length, width, number, arrangement, position, etc.) of each configuration and shape can be changed as appropriate.

  For example, in the above-described embodiment, the calculation process for the case where the inspection target surface is based on black and an abnormal region that is white occurs on the inspection target surface is exemplified, but the present invention is limited to this. First, it can be applied to inspection surfaces and abnormalities of various colors and materials.

  Further, the adoption determining unit 15 (determining unit) determines the pixel value of the master data MD or the updated master data RMD regardless of the similarity between the areas of the master data MD or the updated master data RMD and the matching data FD. Based on at least one of the pixel values of the matching data FD and the matching data FD, it can be determined whether or not the pixel is corrected by the matching data FD. For example, when the inspection target surface is based on white and there is a possibility that an abnormal region that becomes black occurs on the inspection target surface, the adoption determination unit 15 (determination unit) determines that the pixel value of the conformance data FD is When the value is equal to or less than the predetermined threshold value, it can be determined that the value of the pixel is not used for correction. In addition, when the pixel value of the master data MD is equal to or less than a predetermined threshold, the adoption determining unit 15 (determining unit) can determine that the pixel value in the matching data FD is used for correction. In addition, regarding whether correction is possible, various determination criteria based on single or combination of pixel values can be set.

  DESCRIPTION OF SYMBOLS 1 ... Master image data creation apparatus, 13 ... Conformity data creation part (deformation image data creation part), 15 ... Adoption judgment part (judgment part), 16 ... Master data update part (correction part), MD ... Master data (1st Master image data), RMD ... update master data (first master image data), BD ... base data (second master image data), FD ... compatible data (deformed image data).

Claims (7)

By fitting processing based on feature point detection for the first master image data and the second master image data, deformed image data is generated by deforming the second master image data in association with the first master image data. A deformed image data creation unit,
A correction unit that corrects the first master image data based on the deformed image data;
A master image data creation device.   2. A determination unit that determines, for each pixel, whether or not to perform correction by the correction unit based on a pixel value of the first master image data or a pixel value of the modified image data. The master image data creation device described in 1.   The master image data creation device according to claim 2, wherein the determination unit determines whether or not correction is performed by the correction unit based on a similarity between regions in which each pixel is a representative pixel.   4. The master image according to claim 2, wherein the determination unit determines whether to perform correction by the correction unit based on a determination criterion that differs depending on a pixel value of the first master image data. 5. Data creation device.   5. The determination unit according to claim 2, wherein the determination unit determines whether to perform correction by the correction unit based on a determination criterion that differs depending on a pixel value of the deformed image data. Master image data creation device.   Whether the determination unit performs correction by the correction unit based on a determination criterion set according to a combination of a pixel value of the first master image data and a pixel value of the modified image data The master image data creation device according to claim 2 or 3, wherein: The deformed image data creating unit creates a plurality of different deformed image data from one second master image data,
The master image data creation device according to claim 1, wherein the correction unit corrects the first master image data based on the plurality of deformed image data.

Images