TW201415010A - Inspection device, inspection method, and inspection program - Google Patents

Abstract

To provide an inspection device capable of reducing the load for an adjusting operation for inspections when the type of object to be inspected for defects changes or the location of a camera for the object to be inspected changes. [Solution] This inspection device comprises: a registered image storage means for storing registered images, which are images of an object to be inspected, that is from objects to be inspected that are flat, and that is without defects, and for storing a plurality of feature points, which are extracted from the registered image and which are points satisfying prescribed conditions; an alignment means for performing alignment that extracts a plurality of the feature points from an inspection image, which is an image of the object to be inspected captured from a direction different than the capture direction of the captured registered image, that calculates, for a plurality of points on the object to be inspected, a conversion parameter, representing a projection transformation that matches the coordinates of a figure at the point on one of the inspection image and the registered image to coordinates of the figure at the point on the other image, where the figure is contained in both the inspection image and the registered image based on a plurality of the feature points in the inspection image and a plurality of the feature points in the registered image, and that performs the projection transformation on one of the images using the conversion parameter; and a discrepancy detection means for detecting discrepancies in the inspection image and the registered image after the alignment.

Description

Inspection device, inspection method and inspection program

The present invention relates to an inspection apparatus, an inspection method, and an inspection program. More particularly, the present invention relates to an inspection apparatus, an inspection method, and an inspection program for inspecting images.

Investigate whether the text printed on the paper or part is correctly printed, or whether the part is correctly mounted on the substrate, and most of the images obtained by taking the inspection object by the camera are used. This inspection is performed by using the image of the inspection object to determine whether or not the inspection object is defective (that is, whether the inspection object is completed by design). As a method of determining the difference, for example, there is a method of detecting a difference by comparing an image of a non-defective inspection target registered in advance with an image obtained from an inspection target determined to have a defect. In order to detect the difference with high precision by this method, the image acquisition environment for registering the image of the non-defective inspection target and the image acquisition environment for acquiring the image of the inspection target with or without the defect must be the same. The point is quite important. The image acquisition environment is, for example, a camera and a light source. The image acquisition environment is the same, for example, when shooting such images, the camera and camera parameters, the light source, and the positional relationship between the camera and the light source are the same. When the images are acquired separately, the configuration of the inspection object relative to the image acquisition environment must be the same. In order to realize such an inspection environment on the production line, it is necessary to construct an inspection environment in the production line. It takes time or labor to perform an operation of photographing or storing an image of a non-defective inspection object or a reference for detecting a defect from the detected difference. On the other hand, after the construction of the inspection environment is completed, the inspection performed in the inspection environment is much faster than the inspection performed by the human power.

When the production line is less than a large number of production lines, once the aforementioned adjustment is completed, there is no need to wait for a long time. Adjustment. At this time, the time for re-adjustment becomes unnecessary, so the time required to inspect each inspection object is much shorter than the time required to inspect the object one by one. On the other hand, when the production line is added to a small number of various types of production lines in recent years, the inspection environment must be adjusted whenever the type of inspection object is changed. At this time, the inspection efficiency by human inspection may sometimes be high. However, in the visual inspection by manpower, there may be errors that are difficult to avoid completely due to negligence or misunderstanding. In the visual inspection by human power, there is a possibility that the judgment criteria of each examiner are different, resulting in a situation in which quality is unstable. Therefore, the industry hopes to provide an inspection environment that does not require a lot of adjustment work even in a small number of different types of production lines.

Patent Document 1 describes an appearance inspection device that performs a defect inspection by comparing an image of an object to be inspected which is a comparison object with an image of an inspection object to be inspected. In the visual inspection device of Patent Document 1, two or more corresponding pattern portions different from the image of the inspection object to be inspected are imaged, and the two images are aligned. In other words, the visual inspection device sets the position of the region on the image captured by the reference image cut out from the two or more images of the wafer to be compared, that is, the region on the image captured by the wafer of the object to be inspected, in sub-pixel units. detected. Further, the visual inspection device superimposes two images, and the reference image on the image of the wafer to be compared is superimposed on the detected position on the image of the wafer to be inspected. Further, the visual inspection device performs defect inspection by comparing the patterns. Specifically, the visual inspection device generates a difference image of two images, and detects a wafer defect from the generated difference image.

The pattern inspection device of Patent Document 2 detects wafer defects by comparing adjacent wafer images on a semiconductor wafer. When the pattern inspection device compares the adjacent wafer images, the position of the same portion of the adjacent two wafers on the image is deviated and is detected using the image of the portion. The pattern inspection device derives the gray scale difference of the pixels corresponding to each other when the two wafer images overlap based on the detected deviation. The pattern inspection device produces a grayscale difference image as a result of the derivation. The pattern inspection device detects the position of the defect option present on the wafer from the generated grayscale image. The pattern inspection device detects the position of the defect option in the same manner as the image of one of the two wafers that generate the grayscale difference image. The pattern inspection device is located within a predetermined distance from the defect option detected by the previous inspection. When the next inspection also detects the defect option, it is determined that the position of the defect option of the wafer to be inspected is detected by both inspections. Lack of existence trap.

Patent Document 3 describes a surface inspection device that uses two cameras to discriminate between embossing and staining of a steel sheet. Install these two cameras with the beam axis oriented perpendicular to the surface of the steel plate. Further, the surface inspection device photographs the triangular correction piece with two cameras before starting the inspection. The surface inspection device calculates the coordinate transformation parameters of the affine transformation of the coordinate transformation between the two images from the obtained two images. At the time of inspection, the surface inspection device simultaneously photographs the steel plate with two cameras. The surface inspection device performs image conversion by using the coordinate transformation of the coordinate conversion parameters described above, and the coordinates of the two images obtained are the same. The surface inspection device calculates the difference between the pixel values of the respective pixels for the two converted images. The surface inspection device detects the unevenness according to the difference obtained.

[Previous Technical Literature]

[Patent Literature]

[Patent Document 1] Japanese Patent Laid-Open Publication No. 2000-323541

[Patent Document 2] Japanese Patent Laid-Open Publication No. 2004-273850

[Patent Document 3] Japanese Patent Laid-Open Publication No. 2006-177852

In the techniques of Patent Documents 1 to 3, the inspection adjustment operation is required every time the type of the inspection object or the arrangement of the camera with respect to the object is changed. In the techniques of Patent Document 1 and Patent Document 2, for example, the inspection adjustment operation acquires an image for detecting a positional deviation of an object, a part of an image of the object, and an image of the object image without defects. In the technique of Patent Document 3, the adjustment operation for inspection is performed using, for example, a coordinate conversion parameter of a triangular correction piece.

An object of the present invention is to provide an inspection apparatus capable of reducing the burden of an adjustment operation for inspection when the type of the inspection object for detecting a defect or the arrangement of the camera with respect to the inspection object is changed.

The inspection apparatus of the present invention includes: a registered image memory mechanism that memorizes: an image of the inspection object that is not defective in the inspection object in a planar shape, that is, a registered image; and a predetermined condition that is extracted from the registered image Pointing, that is, the feature point; the alignment mechanism performs alignment, and the image is captured from a direction different from the direction in which the registered image is captured, that is, the image is inspected, and the plurality of feature points are extracted; a transformation parameter indicating a projective transformation, the projection transformation is based on the plurality of feature points of the inspection image and the plurality of feature points of the registration image, and for a plurality of points on the inspection object included in both the inspection image and the registration image And causing the coordinates of the image of the inspection image and the image of the registration image to coincide with the coordinates of the image of the point on the other image; and performing the projective transformation on the image by using the transformation parameter; and the difference The detecting mechanism detects the difference between the inspection image and the registered image after the alignment.

In the inspection method of the present invention, the image of the inspection target image which is captured in the inspection object of the plurality of planes, that is, the registration image; and the feature point extracted from the registration image that meets the predetermined condition, that is, the feature point, is memorized. Aligning the image memory mechanism; performing alignment: capturing an image of the inspection object from a direction different from a shooting direction in which the registered image is captured, that is, checking the image, extracting the plurality of feature points; and calculating a projection transformation Converting a parameter, the projective transformation is based on the plurality of feature points of the inspection image and the plurality of feature points of the registered image, and the inspection image is made for a plurality of points on the inspection object included in both the inspection image and the registration image And a coordinate of the image of the point on one of the registered images is consistent with a coordinate of the image of the point on the other image; and performing the projective transformation on the image by using the transformation parameter; and detecting the alignment Check the difference between the image and the registered image.

The inspection program of the present invention enables the computer to implement the following mechanisms: Registering the image memory mechanism, remembering that the image of the inspection object that is not defective in the shape of the plurality of inspection objects, that is, the registration image; and the feature point extracted from the registration image that meets the predetermined condition, that is, the feature point; Aligning the alignment, the aligning includes: capturing an image of the inspection object from a direction different from a shooting direction in which the registered image is captured, that is, checking the image, extracting the plurality of feature points; and calculating a transformation parameter indicating a projective transformation And the projective transformation is based on the plurality of feature points of the inspection image and the plurality of feature points of the registered image, and the inspection image and the plurality of points included in the inspection object included in the inspection image and the registration image Registering the coordinates of the image of the image on one of the images, conforming to the coordinates of the image of the point on the other image; and performing the projective transformation on the image with the transformation parameter; and detecting the alignment by the difference detecting mechanism Then check the difference between the image and the registered image.

Furthermore, the present invention can also be realized by a computer-readable tangible recording medium storing the above-described inspection program.

According to the present invention, when the type of the inspection object or the arrangement between the camera for acquiring the inspection image and the inspection object is often changed, the burden of the inspection adjustment can be reduced, and the inspection including the inspection adjustment time can be shortened. time.

1‧‧‧Checking device

2‧‧‧Photographing device

3‧‧‧Output device

4‧‧‧Check objects

10‧‧‧ Input Department

11‧‧‧Registered Image Memory Department

12‧‧‧Alignment Department

13‧‧‧Differential Detection Department

14‧‧‧Output Department

40‧‧‧Checkpoint

100‧‧‧Check system

1000‧‧‧ computer

1001‧‧‧ processor

1002‧‧‧ memory

1003‧‧‧ memory device

1004‧‧‧I/O interface

1005‧‧‧Recording media

Fig. 1 is a block diagram showing the configuration of an inspection system 100 according to the first, second, and third embodiments.

Fig. 2 is a flow chart showing the operation of the inspection apparatus 1 of each embodiment when the registered image is stored in the registered image storage unit 11.

Fig. 3 is a flow chart showing the operation of detecting the defect of the inspection target 4 by the inspection image device 1 of the first embodiment.

Fig. 4 is a view showing an example of an inspection image.

Fig. 5 is a diagram showing an example of registered images.

Fig. 6 is a flow chart showing the operation of the inspection apparatus 1 at the time of alignment.

Fig. 7 is a view showing an example of feature points extracted from the inspection image and the registration image.

Fig. 8 is a view showing an example of an inspection image before and after the deformation.

FIG. 9 is a view showing an example in which a registered image and a deformed inspection image are detected as defects.

FIG. 10 is a flowchart showing an operation of detecting the defect of the inspection target 4 by the inspection image of the inspection apparatus 1 of the second embodiment.

Fig. 11 is a block diagram showing the correction of the inspection apparatus 1 of the fourth embodiment.

Fig. 12 is a layout view showing the imaging device 2 and the inspection object 4 in the configuration example of the first embodiment.

Fig. 13 is a view showing an example of the detection result of detecting a defect by a mask.

Fig. 14 is a view showing an example of a detection result of detecting a defect by a mask.

FIG. 15 is a view showing an example of the configuration of the computer 1000 used to implement the inspection apparatus 1.

Next, an embodiment of the present invention will be described in detail with reference to the drawings.

Fig. 1 is a block diagram showing the configuration of an inspection system 100 according to a first embodiment of the present invention.

Referring to Fig. 1, the inspection system 100 of the present embodiment includes an inspection device 1, an imaging device 2, and an output device 3. The inspection system 100 detects defects of the inspection object 4.

The inspection device 1 includes an input unit 10, a registered image storage unit 11, an alignment unit 12, a difference detection unit 13, and an output unit 14.

The photographing device 2 is, for example, a camera that captures an image of the inspection target 4.

The output device 3 is a display device such as a display, and may be another device such as a printing device, as long as it is a general output device in a computer system.

The input unit 10 acquires the captured image of the inspection target 4 from the imaging device 2. Hereinafter, the image captured by the object to be inspected, that is, the inspection object 4, is referred to as an inspection image. The input unit 10 can also capture a captured image of the inspection target 4, which is recorded in advance by the photographing device 2, and stored in a registered image storage unit 11 or another memory unit (not shown), as an inspection image.

The image memory unit 11 is registered, and the image of the inspection object 4 without defects is stored. Hereinafter, an image that is compared with an inspection image for detecting a defect is referred to as a registered image. The image of the non-defective inspection object 4 stored in the registered image storage unit 11 is used as a registered image.

The aligning unit 12 calculates a conversion parameter indicating a transformation in which one of the registered image and the acquired inspection image is most consistent with the other. The aligning unit 12 performs image conversion using the calculated transformation parameters.

The difference detecting unit 13 detects a difference between the registered image after the image conversion and the detected image.

The output unit 14 outputs the detected difference to the output device 3.

Next, the operation of the inspection apparatus 1 when the registered image is stored in the registered image storage unit 11 will be described in detail with reference to the drawings.

Fig. 2 is a flow chart showing the operation of the inspection apparatus 1 of the present embodiment when the registered image is stored in the registered image storage unit 11. Hereinafter, storing the registered image in the registered image storage unit 11 is also referred to as registering "registered video". The operation of the inspection apparatus 1 of the other embodiment to be described later when storing the registered image in the registered image storage unit 11 is the same as that shown in FIG. 2.

Referring to Fig. 2, first, the input unit 10 acquires a registered image from the imaging device 2 (step S101).

The photographing device 2 may photograph an image of the inspection object 4 that has been confirmed to have no defect in advance. The method of confirming that the inspection object 4 is free of defects may be any known method. For example, the inspector, by visual inspection, It is confirmed that the inspection object 4 has no defects. Alternatively, the following method may be employed: First, the photographing device 2 photographs three or more different inspection objects 4. The input unit 10 acquires an image captured by the imaging device 2. The alignment unit 12 and the difference detecting unit 13 detect the difference by using one of the combinations of the two images in the captured image as the registered image and the other as the inspection image, and performing the operation described later. The difference detecting unit 13 does not detect the difference in the image combination of the non-defective inspection object 4. On the other hand, the difference detecting unit 13 detects the difference between the image of the defective inspection object 4 and the image of the inspection object 4 without the defect. When the images of the defective inspection object 4 are combined, if the defects are the same, the difference detecting unit 13 does not detect the difference; if the defects are different, the difference detecting unit 13 detects the difference. If the number of non-defective inspection objects 4 is larger than the number of defective inspection objects 4, the number of image detection differences for the non-defective inspection object 4 should be less than the number of image detection differences for the defective inspection object 4. . The input unit 10 acquires the detection result of the difference from the difference detecting unit 13, and detects that the image having the least number of differences is the registered image.

Next, the input unit 10 stores the registered image in the registered image storage unit 11 (step S102).

Next, the operation of the inspection apparatus 1 will be described in detail when the defect of the inspection object 4 is detected by the inspection image with reference to the drawings.

Fig. 3 is a flow chart showing the operation of the inspection apparatus 1 of the present embodiment when detecting defects of the inspection object 4 by inspection images.

Referring to Fig. 3, first, the input unit 10 acquires an inspection image (step S103). As described above, the inspection image is an object to be inspected for the presence or absence of a defect, that is, an image taken by the inspection object 4.

Next, the aligning unit 12 reads the registered image from the registered image storage unit 11 (step S104). As described above, the image taken by the non-defective inspection object 4 is used as the registered image. When the registration image storage unit 11 registers a registered image of a plurality of types of inspection objects 4, for example, the examiner may specify a registration image for the inspection image. Alternatively, the alignment unit 12 may determine the type of the inspection object 4 that is reflected on the inspection image by any known image recognition method. Further, the alignment unit 12 can read the registered image of the inspection object 4 of the determined type from the registered image storage unit 11.

The alignment unit 12 aligns the acquired inspection image with the read registration image (step S105). This alignment corrects the size or shape of the inspection object 4 on the inspection image in accordance with the size or shape of the inspection object 4 on the registered image. The alignment unit 12 performs, for example, a correction for changing the inspection image based on the registered image, thereby performing alignment. In the alignment, the alignment unit 12 calculates the difference between the size and position or direction of the image to be inspected by the image to be erased, or the transformation parameter of the transformation of the inspection object 4 due to the difference in the imaging direction. . These differences are caused by the difference in the type of the photographing device 2 when the registered image is taken, and the positional relationship between the photographing object 4 and the photographing device 2. Further, the alignment unit 12 corrects the inspection image by using the conversion of the calculated transformation parameters. The alignment can also change the registered image according to the inspection image. The alignment will be described in detail later.

Next, the difference detecting unit 13 detects a difference between the aligned inspection image and the registered image (step S106). The detection of the difference will be described in detail later.

Finally, the output unit 14 outputs the difference to the output device 3 (step S107).

Next, the operation of the inspection apparatus 1 at the time of the alignment in step S105 will be described in detail with reference to the drawings.

Fig. 4 is an example of a registered image.

Fig. 5 is an example of an inspection image. The registered image of Fig. 4 is used for the registered image of the inspection image of Fig. 5.

The characters in Fig. 4 and Fig. 5 are the characters on the surface of the inspection object 4. In the present embodiment, these characters are present, and the surface of the inspection object 4 is flat. The text of Fig. 4 and the text of Fig. 5 exist on the same plane, respectively. The defect detected by the inspection apparatus 1 of the present embodiment is a difference in characters on the surface of the inspection object 4. Comparing Fig. 4 with Fig. 5, the portion corresponding to "8" in the lower right character string "abc812" of Fig. 5 is different from Fig. 4. The portion corresponding to "8" in Fig. 5 is "0" in Fig. 4. In the present embodiment, the portion of "8" of the character string "abc812" in Fig. 5 is a defect.

In the two images of the same plane, the relationship between the same point on the plane of the object, that is, the relationship between the coordinates of the two points, is represented by the coordinates of the points on one image and the coordinates of the coordinates of the points on the other image. . The mathematical expression representing the projective transformation, where the coordinates on one image are x1 and the coordinates on the other image are x2, x2 = Hx1. The x1 and x2 are the non-zero 3-dimensional vectors of the third element. The two elements of the vectors are divided by the two values of the first element and the second element, and the two-dimensional vector represented by the two elements represents the coordinates on the plane. The matrix H represents a 3 × 3 matrix of projective transformation. The matrix of the constant multiple of the matrix H also represents the same projective transformation. That is, the degree of freedom of the matrix H is 8. This matrix H representing the projective transformation is also referred to as the homography matrix.

If the homography matrix between two images taken on the same plane is known, it can be calculated at which position of the other image is reflected at all points of one image. Thereby, it is possible to obtain, from an image and a homography matrix, an image obtained by taking a photographing device of one image to capture an image of a subject in another image. In this calculation, it is not necessary to derive the position of the photographing device 2 for taking two images with respect to the subject 4. The photographing device 2 that takes two images can also be different from the same photographing device.

The image of the inspection object 4 on the inspection image obtained by photographing the inspection object 4 without defects is converted into the inspection image by converting the inspection image into the homography matrix of the registration image, and the inspection object 4 on the registered image Like the same. Similarly, the image of the inspection object 4 on the registered image is converted into the registration image by converting the registered image into the homography matrix of the inspection image, and should coincide with the image of the inspection object 4 on the inspection image. Transforming the image with a matrix representing the transformation means transforming the pixels of the image of the transformed object. A description of the specific processing thereof will be described later.

For example, as described in Non-Patent Document 1, it is known that if coordinates of four sets of corresponding points are obtained between two images, the conversion from one image to another can be performed, and the image conversion using the homography matrix (H) can be performed. Export it. That is, if there are four sets of coordinates of the corresponding points in the inspection image and the registered image, the alignment between the inspection image of the size, rotation, and projective deformation and the registered image can be corrected.

(Non-Patent Document 1) Jin Zejing, Jin Gu Jianyi: Automatic Exploration of Feature Points Between Two Images, Image Laboratory, pp. 20-23 (2004)

Fig. 6 is a flow chart showing the operation of the inspection apparatus 1 at the time of alignment.

Referring to Fig. 6, first, the alignment unit 12 extracts feature points from both the inspection image and the registered image (step S301). The alignment unit 12 may also extract feature points from the registered image in advance, and store the coordinates of the extracted feature points in, for example, the registered image storage unit 11. At this time, the alignment unit 12 may read the coordinates of the feature points corresponding to the registered image of the inspection image from the registered image storage unit 11.

The alignment unit 12 extracts the feature points by specifying the coordinates of the pixels that meet the predetermined conditions. For example, the alignment unit 12 may calculate a gradient of pixel values in both the longitudinal direction and the lateral direction of the image. Further, the alignment unit 12 may extract a point at which both the vertical gradient and the horizontal gradient are extreme values as the feature points. The extraction method of the feature points can also be other methods. Further, the alignment unit 12 can extract the feature points by, for example, the method described in Non-Patent Document 2.

(Non-Patent Document 2) David G. Lowe; Object recognition from local scale-invariant features, Proc. of Intl. Conf. on Computer Vision. 2. pp. 1150-1157 (1999).

If the image or the registered image is a character image as shown in FIGS. 4 and 5, the alignment unit 12 may binarize the image with an appropriate threshold value, and obtain a joint region, with the center of gravity of each link region as a feature point. The method extracts feature points. At this time, the aligning unit 12 first generates a pixel value of each pixel as one of two values based on a comparison result between the pixel value of each pixel of the inspection image or the registration image, that is, the original image, and a predetermined threshold value. Value image. For example, when the pixel value of a certain pixel of the original image is above a threshold value, the pixel value of the binary image pixel located at a position corresponding to the pixel position is the first value. When the pixel value of the pixel of the original image is less than the threshold value, the pixel value of the binary image pixel located at the position corresponding to the pixel position is the second value. The first value and the second value are, for example, 0 and 1. The first value and the second value may be any values if they are different values, respectively. Further, the aligning unit 12 extracts a region in which the same pixel value is continuous, that is, a connected region, from the binary image. Check image Or the registered image is a text image, and if the pixel value of the pixel included in the known text corresponds to one of the first value and the second value, the aligning portion 12 only refers to the pixel value of the corresponding text pixel in the binary value. Value, just pull out the link area. Further, the alignment unit 12 may extract the center of gravity of each of the connection regions as a feature point.

Fig. 7 is a view showing an example of feature points extracted from the inspection image and the registration image.

Next, the aligning unit 12 calculates a conversion parameter indicating that the inspection target 4 on the inspection image conforms to the inspection target 4 on the registered image is extracted using the feature points extracted from the two images (step S302).

As described above, when this transformation is represented by a homography matrix, the transformation parameters are the values of the elements of the homography matrix. At this time, the alignment unit 12 calculates a homography matrix, that is, a transformation matrix H. As described above, in order to calculate the transformation matrix H, four or more corresponding point groups are required. Corresponding point group: A group of points on the inspection image and points on the registered image that can be regarded as images of the same point on the subject.

When the transformation matrix H is calculated, the alignment unit 12 first detects four or more corresponding point groups. The alignment unit 12 may detect the corresponding point group by a method of simple template matching or a known method described in Non-Patent Document 1 or the like.

The alignment unit 12 calculates the transformation matrix H based on the coordinates of the corresponding group of four or more detected points. The alignment unit 12 may calculate the transformation matrix H by any known method such as a method of calculating each element of the transformation matrix H by a least square method.

The aligning unit 12 can also calculate the transformation matrix H by "stable estimation of the influence of the deviation value in the coordinates of the corresponding point group". It is detected as a corresponding point group, but it is not actually a coordinate of the "group of points on the inspection image and the points on the registered image" of the corresponding point group. The deviation value of the coordinates of the corresponding point group is caused by, for example, erroneous extraction of feature points or erroneous detection of corresponding point groups. As a method of robust estimation, for example, there is an estimation method based on LMedS (Least Median of Squares) estimation or M-estimator. Alignment portion 12 The transformation matrix H is calculated by, for example, such methods or other types of robust estimates.

The alignment unit 12 may be calculated by the method described in Non-Patent Document 3, for example. The method described in Non-Patent Document 3 calculates a transformation matrix H while correlating feature points extracted from both the inspection image and the registered image.

(Non-Patent Document 3) Martin A. Fischler and Robert C. Bolles; Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography, Comm. of the ACM 24(6), pp.381-395 ( 1981).

The alignment unit 12 can also calculate the transformation matrix H by using the method described in Non-Patent Document 4, for example. This method is a method of calculating a transformation matrix H from the coordinates of the corresponding point group after calculating the corresponding point group after extracting the feature points.

(Non-Patent Document 4) Akiyama and Nishiwaki: Proposal for the use of projection invariants and homography matrices for image search methods, the 9th Information Technology Forum (FIT2010), H-011 (2011.9).

Next, the aligning unit 12 performs conversion of the inspection image using the transformation matrix H (step S304).

When the inspection image is converted, the aligning unit 12 first converts the coordinates of the transformation matrix H to the respective pixels of the inspection image. Next, the aligning unit 12 calculates the pixel values of the converted inspection images in the coordinates of the pixels of the registered image based on the pixel values of the pixels of the inspection image and the transformed coordinates. Alternatively, the second alignment unit 12 may calculate the pixel value of the registered image in the coordinates of each of the pixel elements of the inspection image.

Fig. 8 is a view showing an inspection image before and after the transformation by the transformation matrix H. In the present embodiment, the transformation is also referred to as deformation. According to the present transformation, the difference in the size of the inspection object 4 with respect to the inspection image of the registered image, the difference in the direction of the inspection object 4, and the projective distortion are all eliminated.

As described above, the alignment unit 12 converts the inspection image and checks that the image of the inspection object 4 included in the image conforms to the image of the inspection object 4 included in the registered image. Conversely, the alignment unit 12 can also change the registered image, and the image of the inspection object 4 included in the registration image conforms to the image of the inspection object 4 included in the inspection image.

Next, the difference is detected in detail with respect to step S106: the difference detecting unit 13 detects the difference.

In step S106, the difference detecting unit 13 calculates the "difference between the registered image after the conversion and the inspection image" indicated by the calculated conversion parameter. In the above example, the difference detecting unit 13 calculates a difference between the registered image and the inspection image that is deformed using the calculated deformation matrix H. Specifically, the difference detecting unit 13 may calculate “the difference between the registered image and the pixel value of the inspection image deformed by using the calculated deformation matrix H in the same coordinate”. The difference detecting unit 13 can also calculate the difference between the feature amounts calculated from the pixel values of the two images as the difference between the two images.

The difference detecting unit 13 outputs the detected difference to the output unit 14 as information indicating the detected defect. The difference detecting unit 13 may output, for example, the information of each coordinate included in the designated image or the registered image and the difference calculated, and a value indicating the difference between the coordinates. The coordinates can be coordinates in the check mark before the change, or can be the coordinates in the registered image. The difference detecting unit 13 may specify a coordinate in which the calculated difference exceeds the predetermined threshold S, and output the designated coordinate. The difference detecting unit 13 may output a difference in the coordinates in addition to the designated coordinates. In such a case, the threshold S may be preset by, for example, the designer of the system or an inspector.

In step S107, the output unit 14 outputs information indicating the difference received from the difference detecting unit 13 to the output device 3.

The output unit 14 generates, for example, a pixel value of each pixel indicating a difference image of the difference in the pixel coordinates. Further, the output unit 14 may display the generated difference image on the output device 3. The output unit 14 may change the color or the gradation value of each pixel of the inspection image, for example, according to the difference in the detected difference. Further, the output unit 14 may display the changed inspection image on the output device 3. Alternatively, the output unit 14 may be modified to check the pixel color included in the region where the difference in the image exceeds the threshold S. Or shade value. Further, the output unit 14 may display the changed inspection image on the output device 3. The output unit 14 may be modified to check a pixel color or a gradation value included in, for example, a circumscribed rectangle of a connection region of a region where the difference in the image exceeds the threshold S. Further, the output unit 14 may display the changed inspection image on the output device 3. The output unit 14 may change the color or the gradation value of the registered image without checking the image. Further, the output unit 14 may display the changed registered image on the output device 3. The output unit 14 can also display differences in characters. The output unit 14 can also change the color or the gradation value or the like by applying one or both of the inspection image and the registration image as described above, and can specify the change of the portion where the difference exists. Further, the output unit 14 may display and display the changed inspection image and registered image on the output device 3. The difference form in which the output unit 14 is displayed on the output device 3 is not limited to the above.

As described above, the output device 3 is a general output device in a computer system such as a display or a printer.

Fig. 9 is a view showing a difference between the registered image and the deformed inspection image as a defect detection.

In the present embodiment, when the type of the inspection object or the arrangement between the camera for acquiring the inspection image and the inspection object is often changed, the burden of adjustment for inspection can be reduced, and the adjustment for inspection can be shortened. Time to check the time.

The reason for this is that the alignment unit 12 performs the inspection of the inspection target image 4 on the inspection image or the registration image, and the change in the size, direction, and projective deformation of the inspection object 4 on the inspection image. Further, the difference detecting unit 13 detects the difference between the converted inspection image and the registered image. Therefore, when the registered image is acquired and the inspection image is acquired, the positional relationship between the inspection object 4 and the imaging device 2 is preferably close, but it is not necessary to be the same. Thereby, the arrangement of the photographing device 2 and the inspection object 4 does not need to be closely adjusted. Further, even if the arrangement of the photographing device 2 and the inspection object 4 is changed, it is not necessary to perform correction.

The alignment unit 12 calculates the parameters of the above-described conversion using the coordinates of the feature points extracted from the image of the inspection object 4. Therefore, the inspection object 4 does not need to be used for the alignment mark. There is no special tool to be used for correction. Even if the type of the inspection object 4 is changed, it is not necessary to check the adjustment.

Moreover, the inspection system 100 of the present embodiment can also be used for most of the visual inspection, and a small number of types of inspections. Therefore, by performing the inspection by the inspection system 100, it is possible to suppress the occurrence of a malfunction in the visual inspection performed at the time of a small number of types of inspections, or the quality difference caused by the difference in the inspection basis of each examiner.

Next, a second embodiment of the present invention will be described in detail with reference to the drawings.

The configuration of the inspection system 100 of the present embodiment is the same as the configuration of the inspection system 100 of the first embodiment of Fig. 1 .

The registered image storage unit 11 of the present embodiment stores the registered image of the plurality of types of inspection objects 4.

The alignment unit 12 performs alignment with the inspection image acquired by the imaging device 2 for all registered images stored in the registration image storage unit 11.

The difference detecting unit 13 detects the difference between the registered image and the detected image after the alignment for all the registered images stored in the registered image storage unit 11. The difference detecting section 13 selects the smallest difference to select the difference as a difference with respect to the inspection image.

The constituent elements of the inspection system 100 of the present embodiment are the same as those of the first embodiment, and the description thereof is omitted.

Next, the operation of the inspection apparatus 1 of the present embodiment will be described in detail with reference to the drawings.

Fig. 10 is a flow chart showing the operation of the inspection apparatus 1 of the present embodiment for detecting defects of the inspection object 4 by inspection images.

First, the input unit 10 acquires an inspection image (step S103).

When the registered image stored in the registered image storage unit 11 has a registered image in which the difference between the acquired inspection image and the acquired inspection image is not detected (Yes in step S201), the alignment unit 12 selects one registered image from the registered image. (Step S202).

The alignment unit 12 detects the difference between the selected registered video and the alignment unit 12 of the first embodiment (step S104, step S105, and step S106).

When all the registered images stored in the registered image storage unit 11 are detected and the difference between the registered image and the acquired inspection image is completed (step S201, No), the difference detecting unit 13 selects the smallest difference (step S203).

The difference detecting unit 13 selects the smallest difference, and corresponds to the difference detecting unit 13 selecting the registered image derived from the selected difference as the registered image corresponding to the acquired inspection image.

The output unit 14 outputs the selected difference to the display device 3 (step S107).

The calculation method of the difference between the registered image and the inspection image may be any known method for calculating the evaluation value of the difference between the images. The difference detecting unit 13 calculates, for example, the sum of the absolute values of the pixel values of the registered image pixels and the test image pixels of the coordinates corresponding to the registered image and the image of the inspection image after the conversion. Just fine. The difference detecting unit 13 may have a difference in the calculated sum. The difference detecting unit 13 may be configured such that the difference between the pixel values of the registered image and the inspection image exceeds the threshold value S after the conversion, and the number of the registered image pixels and the inspection image pixel group located at the coordinates corresponding to each other is different.

In the present embodiment, in addition to the effects similar to those of the first embodiment, the effect of reducing the inspection load can be further reduced. Further, in the present embodiment, there is an effect that even if different types of inspection objects 4 are mixed in a random order, it is possible to perform inspection without identifying the type of the inspection object 4 or the adjustment of the inspection object type change. .

The reason is that the difference detecting unit 13 depends on each registered image stored in the registered image storage unit 11. Choose the smallest difference from the difference in the image. The difference derived from the registered image of the inspection object 4 of a different type from the inspection target 4 of the inspection image is generally larger than the difference derived from the registered image of the inspection object 4 of the same type as the inspection target 4 of the inspection image. Therefore, if the registered image of the inspection object 4 of the same type as the inspection target 4 of the inspection image is stored by the registration image storage unit 11, the minimum difference selected by the difference detection unit 13 is the same as the inspection target 4 of the inspection image. Check the difference derived from the registered image of the object 4. Therefore, it is not necessary to specify a registered image when checking. When the type of the inspection object 4 is changed, it is not necessary to perform inspection adjustment.

Next, a third embodiment of the present invention will be described in detail with reference to the drawings.

The configuration of the inspection system 100 of the present embodiment is the same as the configuration of the inspection system 100 of the first embodiment of Fig. 1.

However, the difference detecting unit 13 does not calculate the difference between the registered image and the detected image pixel value after the image conversion, and calculates the feature amount from the image pixel values. Further, the difference detecting unit 13 outputs the difference between the feature amounts as the difference between the registered image and the inspection image to the output unit 14.

Next, the operation of the inspection apparatus 1 of the present embodiment will be described in detail with reference to the drawings.

The inspection object 4 of the present embodiment is a pattern of one or more types of predetermined patterns on a planar surface. As such a figure, for example, there is a text. Such a graphic may also be a graphic other than a circle, a polygon, a combination of line segments, and the like.

Fig. 3 is a flow chart showing the operation of the inspection apparatus 1 of the present embodiment for detecting defects of the inspection object 4 based on the inspection image.

Since the operations of steps S103 to S105 of the inspection apparatus 1 of the present embodiment are the same as those of the inspection apparatus 1 of the first embodiment, the description thereof is omitted.

In step S106, the difference detecting unit 13 extracts both the registered image and the inspected image from the image conversion. The feature quantity indicating the figure is displayed. The difference detecting unit 13 also specifies a coordinate indicating the position of the figure. The feature quantity representing the figure is, for example, a vector representing the shape of the figure. As mentioned above, the graphics are for example text. The difference detecting unit 13 may be, for example, a coordinate of the center of gravity of the detected pattern as a coordinate of the graphic position. For example, the difference detecting unit 13 may detect the coordinates of the center of gravity of the rectangle circumscribing the rectangle, or the coordinates of any of the vertices determined in advance as the coordinate of the pattern position.

The feature amount indicating the pattern on the registered image, or the position coordinate at which the feature amount is extracted may be stored in advance by the registered image storage unit 11. At this time, the difference detecting unit 13 may read the feature amount indicating the pattern on the registered image and the position coordinate detected by the pattern from the registered image storage unit 11.

Next, the difference detecting unit 13 extracts a pattern of a region overlapping the pattern detected from the other image when the registered image and the inspection image overlap each other in the pattern detected from the registered image and the inspection image. The difference detecting unit 13 calculates the ratio of the area of the area overlapping the area of one of the pattern areas with respect to the combination of the pattern detected by the registered image and the pattern detected by the inspection image. The graphic area can also be a rectangle outside the graphic. The difference detecting unit 13 maximizes the ratio of the areas that overlap each other, and the pattern extracted from the registered image is associated with the pattern extracted from the inspection image.

Further, the difference detecting unit 13 calculates the difference between the two pattern feature amounts associated with each other and the coordinate difference between the positions at which the patterns are extracted. If the feature amount is a vector, the difference detecting unit 13 may calculate the distance defined in advance between the two feature amounts as the feature amount difference. The difference detecting unit 13 may calculate, for example, the distance between two points indicated by the position coordinates extracted by the two figures as the coordinate difference. The difference detecting unit 13 may set, for example, a distance between the feature amount difference of the pattern in which the associated figure does not exist and the position coordinate, and set a predetermined value larger than a threshold value to be described later.

The difference detecting unit 13 compares the distances of the two graphic feature amounts associated with each other with respect to the respective patterns, and the threshold value S2 with respect to the feature amount distance. Further, the difference detecting unit 13 compares the distance between the position coordinates detected by the two patterns associated with each other with respect to the respective patterns, and the threshold value S3 with respect to the distance between the coordinates. The difference detecting unit 13 detects, as a defect, a pattern in which at least one of the distance between the feature amount distance and the position coordinates detected by the two patterns exceeds each threshold value.

The difference detecting unit 13 may transmit information indicating a pattern area detected as a defect to the output unit 14, for example. The difference detecting unit 13 may further transmit any of the above-described distances exceeding the threshold to the output unit 14.

Similarly to the first embodiment, the output unit 14 can display the detection result of the defect by specifying the shape at which the defect is detected.

Further, similarly to the detection device 1 of the second embodiment, the detection device 1 of the present embodiment can detect the difference between the registered image and the inspection image for all the registered images stored in the registered image storage unit 11. Further, the detecting device 1 of the present embodiment may output the difference as a defect when the smallest difference detected exceeds the threshold.

This embodiment has the same effects as those of the first embodiment.

Next, a fourth embodiment of the present invention will be described in detail with reference to the drawings.

Fig. 11 is a view showing the configuration of the inspection apparatus 1 of the present embodiment.

The inspection apparatus 1 of the present embodiment includes a registered image storage unit 11 that stores a plurality of images of the inspection target image that are not defective in the inspection target in a plane, that is, a registered image; and the extracted image from the registered image a point that meets a predetermined condition, that is, a feature point; the alignment mechanism 12 performs alignment, and the alignment includes: capturing an image of the inspection object from a direction different from a photographing direction in which the registered image is captured, that is, inspecting the image, and extracting the plurality of a feature point; calculating a transformation parameter indicating a projective transformation, wherein the projection transformation is based on the plurality of feature points of the inspection image and the plurality of feature points of the registered image, and the inspection object is included in both the inspection image and the registration image a plurality of points on the image of the inspection image and the image of the image on the image of the registration image, and a coordinate of the image of the image on the other image; and the projection of the image by the transformation parameter Transformation; and The difference detecting unit 13 detects the difference between the inspection image and the registered image after the alignment.

Fig. 11 is a view showing the configuration of the inspection apparatus 1 of the present embodiment.

Next, a specific configuration example of the first embodiment will be described in detail with reference to the drawings.

The configuration of this configuration example is the same as that of the first embodiment of Fig. 1.

Fig. 12 is a view showing the arrangement of the photographing device 2 and the inspection object 4 of the present configuration example.

Referring to Fig. 12, the imaging device 2 of the present configuration example images the planar inspection object 4 placed on the inspection table 40.

The x direction and the y direction of Fig. 12 are the directions parallel to the surface of the inspection table 40 and the inspection object 4. The z direction is the surface of the inspection table 40 and the normal direction of the inspection object 4. The distance between the inspection object 4 and the image plane projected by the image in the photographing device 2 is d. The off-angle of the optical axis of the photographing device 2 with respect to the surface of the inspection table 40 and the normal direction of the inspection object 4 is represented by θ and γ. As long as the inspection object 4 is included in the depth of field of the photographing device 2, the image of the inspection object 4 in the image captured by the photographing device 2 is in a focused state as a whole, and the distance d, the off angles θ, and γ may be any values.

The inspection device 1 is connected to the photographing device 2. The inspection device 1 and the photographing device 2 may be connected by a general-purpose interface such as USB (Universal Serial Bus) or IEEE 1394 (The Institute of Eelctrical and Electoronics Engineers). The inspection device 1 can also be constructed using a computer equipped with an image capture plate. At this time, the inspection device 1 and the photographing device 2 may be connected by a transmission line that can transmit a video signal. Further, the inspection device 1 can also obtain the image signal transmitted from the photographing device 2 as a digitized image by the image capture panel.

Next, explain the effect when checking the keyboard.

As one of the keyboard inspection methods, a method of using a mask is known. Mask can be used to shade Inspection of the inspection object of the pattern. The mask is a sheet of a pattern in which the pattern of the object is indented and inverted. The inspection of the inspection object using the mask is performed by superimposing the mask on the inspection object. The portion of the mask corresponding to the black color of the inspection object is transparent. Such a mask is made of a vinyl sheet or the like. When the non-defective inspection object overlaps the mask, no white portion appears on the mask. If there is a white defect in the original black area of the inspection object, when the inspection object overlaps the mask, a white portion appears on the mask. The inspector will detect the white portion appearing due to the overlap of the inspection object as a defect.

For example, the mask used to check the background color is white and the text color is black. The background color is black, and the text is partially transparent.

Fig. 13 is a view showing an example of inspection results of detecting defects by a mask. FIG. 13 shows the result of inspection of the keyboard pattern, the mask, and the overlay mask as the inspection object on the keyboard.

In the keyboard of Fig. 13, there is a defect in printing V where the printing should be performed. Because of this keyboard overlap mask, the portion in which W is printed as V is white in a pattern in which white W overlaps black V. The inspector will display the white figure as a defect detection.

However, defects existing in the portion covered by the mask when the mask is overlapped may sometimes detect defects without mask detection.

Fig. 14 is a view showing an example of an inspection result of detecting a defect by a mask. In the keyboard shown in Fig. 14, there is a defect in printing O at the printing C. In the case of the example shown in Fig. 14, even if the keyboard overlaps the mask, no white pattern appears in the presence of the defect. Therefore, the examiner cannot detect the defect. When printing E by mistake in the keyboard printing F, the defect cannot be detected by the mask inspection.

The inspection apparatus 1 according to each embodiment of the present invention detects a difference in the comparison target image. Therefore, the inspection apparatus 1 according to each embodiment of the present invention can detect the aforementioned defect that cannot be detected by the mask mask during the mask inspection.

The inspection device 1 can be realized by a computer and a program for controlling the computer, a dedicated hardware, or a combination of a computer and a program for controlling the computer and a dedicated hardware.

FIG. 15 is a view showing an example of the configuration of the computer 1000 used to implement the inspection apparatus 1. Referring to FIG. 15, the computer 1000 includes a processor 1001, a memory 1002, a memory device 1003, and an I/O (Input/Output) interface 1004. And the computer 1000 can access the recording medium 1005. The memory 1002 and the memory device 1003 are recording devices such as a RAM (Random Access Memory) and a hard disk. The recording medium 1005 is a memory device such as a RAM or a hard disk, a ROM (Read Only Memory), or a portable recording medium. The memory device 1003 may also be the recording medium 1005. The processor 1001 can read and write data or programs to the memory 1002 and the memory device 1003. The processor 1001 can receive, for example, the image data of the inspection object by the photographing device 2 via the I/O interface 1004. The processor 1001 can access the recording medium 1005. In the recording medium 1005, a program for causing the computer 1000 to operate as the inspection device 1 is stored.

The processor 1001 stores the program stored in the recording medium 1005 and causes the computer 1000 to operate as the inspection device 1 in the memory 1002. Further, since the processor 1001 executes the program loaded in the memory 1002, the computer 1000 operates as the inspection device 1.

The input unit 10, the aligning unit 12, the difference detecting unit 13, and the output unit 14 can be read to the memory 1002 by, for example, the recording medium 1005 from the memory program, and a dedicated program for realizing each function can be executed. The device 1001 is implemented. The registered image storage unit 11 can be realized by the memory 1002 included in the computer 1000 or the memory device 1003 as a hard disk device. Alternatively, part or all of the input unit 10, the registered image storage unit 11, the alignment unit 12, the difference detecting unit 13, and the output unit 14 may be realized by a dedicated circuit that realizes the functions of the respective units.

The present invention has been described above with reference to the embodiments, but the present invention is not limited by the embodiments described above. Various modifications can be made without departing from the scope of the invention.

This application claims a special request for a Japanese application filed on September 18, 2012. The priority based on 2012-204834 is hereby incorporated by reference.

[Industrial use]

According to the present invention, it is applicable to inspection of a plane on which characters, figures, and the like are printed. The invention can also be applied to configuration check of shortcut keys such as a keyboard.

1‧‧‧Checking device

2‧‧‧Photographing device

3‧‧‧Output device

4‧‧‧Check objects

10‧‧‧ Input Department

11‧‧‧Registered Image Memory Department

12‧‧‧Alignment Department

13‧‧‧Differential Detection Department

14‧‧‧Output Department

100‧‧‧Check system

Claims (10)

An inspection device comprising: a registered image memory mechanism, remembering: an image of the inspection object that is not defective in a plurality of shapes in a planar inspection object, that is, a registered image; and a predetermined condition that is extracted from the registered image a point, that is, a feature point; the alignment mechanism performs alignment, the image includes: capturing an image of the inspection object from a direction different from a shooting direction in which the registered image is captured, that is, checking the image, extracting the plurality of feature points; and calculating the representation a projective transformation conversion parameter, the projective transformation is based on the plurality of feature points of the inspection image and the plurality of feature points of the registration image, and for a plurality of points on the inspection object included in both the inspection image and the registration image, And making a coordinate of the image of the inspection image and the image of the registration image coincide with a coordinate of the image of the point on the other image; and performing the projective transformation on the image by using the transformation parameter; and detecting the difference The mechanism detects the difference between the inspection image and the registered image after the alignment. The inspection device of claim 1, wherein the registered image memory device stores a plurality of the registered images of the registered image including the one or more types of the inspection object, and the plurality of feature point groups of the registered image, the alignment The mechanism detects the registration image and the inspection image for each of the plurality of registered images stored by the registered image storage means, and the difference detecting means detects the difference between the aligned image and the inspection image after the alignment, Among the differences detected, the difference indicating the smallest value of the difference is detected as the difference with respect to the inspection image. The inspection apparatus according to claim 1 or 2, wherein the difference detecting means calculates a feature amount from at least a part of the inspection image and the registered image area of the corresponding area, and calculates a difference between the calculated feature amounts as the The difference was detected. The inspection device of claim 1, wherein the registered image memory device memorizes at least a predetermined number of features extracted from the registered image a coordinate of the point, the alignment mechanism extracts at least a predetermined number of the feature points from the inspection image, and a plurality of feature points corresponding to at least a predetermined number detected from the inspection image and one of the registration images, The feature point extracted by one image is detected, and the transformation parameter is calculated according to a complex combination of the feature point extracted from the inspection image and the feature point extracted from the registration image. The inspection apparatus of claim 4, wherein the alignment mechanism uses a robust estimate of coordinates of the feature point that mitigates the influence of the deviation value on the transformation parameter in the coordinates of the complex feature point, Calculate the transformation parameters. The inspection apparatus of claim 1, wherein the alignment mechanism generates a binary image in which the inspection image is binarized with a predetermined threshold, and the pixel value of one or both of the binary images is extracted as having The connected region of the continuous region of the pixel value is extracted as the feature point. An inspection system comprising: a photographing device for photographing the inspection image; and an inspection device according to claim 1 of the patent application and a display device for receiving and displaying the difference from the inspection device. An inspection method for recording an image of the inspection object that is not defective in the inspection object of a plurality of planes, that is, a registered image; and a feature point extracted from the registration image that meets a predetermined condition, that is, a feature point, is stored in the registration The image memory mechanism performs alignment: the image is captured from a direction different from the direction in which the registered image is captured, that is, the image is inspected, and the plurality of feature points are extracted; and a transformation parameter indicating a projective transformation is calculated. , the projective transformation is based on the plural of the inspection image The feature point and the plurality of feature points of the registered image are such that the image of the check image and the image of the registered image is imaged at a plurality of points on the inspection object included in the inspection image and the registration image a coordinate that coincides with a coordinate of the image of the point on the other image; and performs the projective transformation on the image by the transformation parameter; and detects a difference between the inspection image and the registration image after the alignment. A computer-readable recording medium recording an inspection program for causing a computer to implement an image recording memory mechanism that memorizes: an image of the inspection object that is not defective in a plurality of inspection objects in a planar shape Registering an image; and aligning with a point extracted from the registered image that meets a predetermined condition, that is, a feature point; the alignment includes: taking the inspection from a direction different from a shooting direction in which the registered image is taken An image of the object, that is, an inspection image, extracting the plurality of feature points; and calculating a transformation parameter indicating a projective transformation, wherein the projection transformation is based on the plurality of feature points of the inspection image and the plurality of feature points of the registration image, and the image is included in the image Checking a plurality of points on the inspection object of the image and the registered image such that the coordinates of the image of the point on the inspection image and the one of the registration images are consistent with the coordinates of the image of the point on the other image; and Performing the projective transformation on the one image by using the transformation parameter; and detecting the shadow after the alignment The difference of the registered image. The computer-readable recording medium of claim 9, wherein the inspection program causes the computer to: the registered image memory mechanism memorizes the plurality of registered images of the registered image including the type of the inspection object; a plurality of the feature point groups of the registration image; the alignment mechanism aligning the registration image with the inspection image for each of the plurality of registered images stored by the registration image storage mechanism; and the difference detecting mechanism for the pair The difference between the registered image and the inspection image is detected, and the difference is detected. The difference indicating the smallest value of the difference is detected as the difference to the inspection image.