JP2012058197A - Generation method for test image, image test method using the same, and appearance test device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform an image test of a product to test exhaustively even when the product to test is a flexible packaging container of considerable flexibility.SOLUTION: Division cells in which a reference image is divided into a two-dimensional matrix are set. Among the division cells, those which include coordinates representing patterns characteristics to be tested as a reference point are set as correction application cells. A picked-up image obtained by being picked up and the reference image are collated by pattern matching to detect a shift amount of the picked-up image corresponding to the correction application cell. On the basis of the shift amount detected, the shift amounts of all the division cells other than the correction application cell in the picked-up image are operated by linear interpolation. On the basis of an operation result of the linear interpolation, a test image is generated. The generated test image and the reference image are collated to determine whether or not the product is good.

Description

  The present invention relates to an inspection image generation method, an image inspection method using the inspection image, and an appearance inspection apparatus.

  A technique is known in which a reference image is stored in advance and an inspection image obtained by imaging a workpiece is compared with a reference image to determine whether the workpiece is acceptable or not. In addition, a technique for obtaining an inspection image by correcting distortion of a captured image obtained by imaging prior to collation is also known. For example, Patent Document 1 discloses a technique for inspecting a positional shift sequentially in units of line blocks obtained by dividing an image for each horizontal scanning line, aligning the reference image, and comparing the reference image with the inspection image. Further, Patent Document 2 discloses a description in which a plurality of surfaces of a workpiece are imaged with a plurality of cameras, and after these are combined to obtain an inspection image, the inspection image and the reference image are compared.

Japanese Patent No. 3725207 JP 2000-171409 A

  In spite of the various processes as described above, there are cases where sufficient image correction cannot be achieved depending on the mode of imaging and the type of product.

  As an example, in recent years, there has been a demand for visual inspection of soft packaging material containers having a high peripheral surface flexibility, such as cosmetic tubes and seasoning containers. This appearance inspection requires a step of imaging the entire outer periphery of the soft packaging material container and comparing the developed image as an inspection image with a reference image.

  However, it is not easy to precisely image the entire outer periphery of the soft packaging material container.

  In imaging, it is usually preferable to insert a mandrel or other jig into the inner periphery of the inspection object and rotate the inspection object around its center line to image the surface. However, the inspection object inserted into the jig is not a true circle. Therefore, the peripheral speed of the product to be inspected varies during imaging. This variation is different for each product to be inspected, and the degree is different for each imaging. As a result, the image is distorted, and a non-defective product may be erroneously determined as a defective product.

  In the configuration of Patent Document 1, the relative positional deviation between the inspection image and the reference image is corrected for each line block. However, for the cells constituting each line block, the deviation of the selected cell applied exclusively as a template is corrected. The minimum value of the quantity is only averaged, and the deviation amount in all the cells constituting the line block cannot be reflected. For this reason, it has been impossible to correct the distortion generated in the inspection image with a required accuracy when the peripheral surface of the rotating inspection target product is imaged.

  The configuration of Patent Document 2 also cannot correct distortion generated in the inspection image. In addition, a configuration using a plurality of cameras causes high costs.

  The present invention has been made in view of the above-described problems. Even when the inspection target product is a highly flexible soft packaging material container, an inspection image capable of precisely performing image inspection of the inspection target product. It is an object to provide a generation method, an image inspection method using the generation method, and an appearance inspection apparatus.

  In order to solve the above-described problems, the present invention captures an image of an inspection object relative to an imaging device, and inspects an inspection image that is compared with a reference image that is a reference for pass / fail of the inspection object from the obtained captured image. In the inspection image generation method for generating the reference image, a divided cell setting step for setting a divided cell obtained by dividing the reference image into a two-dimensional matrix, and coordinates representing the characteristics of the pattern to be inspected among the divided cells are set as reference points. A correction application cell setting step for setting a cell to be included as a correction application cell, pattern matching between the captured image and the reference image, and a deviation amount of the captured image corresponding to the correction application cell for each reference point. Based on the amount of deviation detected at the reference point detected in the amount of deviation detection step and the amount of deviation at the reference point detected in the amount of deviation detection step. An interpolation calculation step for calculating the deviation amount of all the divided cells by linear interpolation, and an inspection for generating an inspection image in which the captured image is corrected for each divided cell based on the deviation amount of all the divided cells including the correction application cell An inspection image generation method characterized by comprising an image generation step. In this aspect, since the shift amount of the captured image is calculated for all the divided cells by interpolation based on the shift amount obtained by pattern matching between the captured image and the reference image, the speed at the time of imaging of the inspection target product Even when distortion due to the variation in the image and other causes occurs at the time of imaging, a precise inspection image in which the distortion is corrected for every divided cell can be obtained. As a result, the accuracy of the inspection image is improved, and erroneous determination at the time of collation can be reduced as much as possible.

  In a preferred aspect, a polygon setting step for setting a polygon divided into triangles having the reference point as a vertex, and any one of the divided cells other than the correction application cell as a calculation target for linear interpolation. A normal line that calculates a normal vector along the normal direction of the divided cell of the captured image with a scalar quantity corresponding to the deviation amount for each vertex of the polygon. A vector setting step, and a matrix calculation step for calculating, for each component of the normal vector, the shift amounts of all the divided cells assigned to the polygon based on the calculated end point of the normal vector. . In this aspect, when calculating linear interpolation, a determinant can be set based on the end point of a normal vector with the vertex of a triangular polygon as a base point, and the amount of deviation can be calculated by linear interpolation with a simple algebra.

  In a preferred aspect, there is provided a forcible setting step for forcibly setting the divided cells contributing to the generation of the polygons as correction application cells. In this aspect, when a polygon is set, it is possible to prevent the area from becoming excessively large or to prevent generation of divided cells that are not included in the polygon, thereby suppressing a calculation error of a suitable deviation amount. .

  In a preferred embodiment, an expansion step of continuing the same reference image on both sides of the reference image, a divided cell corresponding to the correction application cell among the reference images continuous on both sides, and a reference image continuous on both sides Registering the corner divided cells as correction application cells. In this aspect, all the divided cells can be assigned to the polygon, and the deviation amount for each divided cell can be reliably calculated by linear interpolation.

  In a preferred aspect, the polygon setting step includes a step of dividing the polygonal polygon into triangular polygons by a two-dimensional Delaunay division method. In this aspect, it is possible to perform a suitable three-dimensional mapping that can cover each correction application cell as a vertex.

  In a preferred aspect, the matrix calculation step includes a start point of an end point of one normal vector calculated from the vertex of the polygon as a base point, and another normal line calculated from the other vertex constituting the polygon as a base point. Two unit vectors toward the end point of the vector and a determinant of the coefficient based on the component vector of each unit vector are calculated in advance, and for each divided cell assigned to the polygon based on these unit vectors and the determinant A step of sequentially calculating the amount of deviation; In this aspect, since the unit cell and determinant necessary for calculating the deviation amount are sequentially calculated in a state in which the unit vector and determinant are calculated in advance, it is possible to eliminate duplication processing and improve the efficiency of the calculation. it can.

  According to another aspect of the present invention, in the image inspection method using the above-described inspection image generation method, the inspection image generated by the inspection image generation method is compared with the reference image, and the inspection image passes or fails. An image inspection method comprising a pass / fail determination step for determining whether or not.

  Another aspect of the present invention is an imaging device that images an inspection site of a product to be inspected, a reference image storage unit that stores a reference image serving as a reference for pass / fail of the product to be inspected, and an inspection target among the reference images Among the divided cells, a reference point storage unit that stores coordinates representing the features of the pattern as a reference point, a divided cell information storage unit that stores information about divided cells obtained by dividing the reference image into a two-dimensional matrix, A correction application cell storage means for storing a cell including the reference point as a correction application cell, a captured image captured by the imaging apparatus and the reference image are collated by pattern matching, and the captured image corresponding to the correction application cell Based on the deviation amount at the reference point detected in the deviation amount detection step, and the deviation amount detection means for detecting the deviation amount for each reference point. Interpolation calculation means for calculating a deviation amount of all divided cells other than the normal application cell by linear interpolation, and an inspection for generating an inspection image in which the captured image is corrected for every divided cell based on the deviation amount of each divided cell An appearance inspection apparatus comprising: an image generation unit; and a determination unit that compares the inspection image with the reference image to determine whether the inspection target product is good or bad. In this aspect, since the shift amount of the captured image is calculated for all the divided cells by interpolation based on the shift amount obtained by pattern matching between the captured image and the reference image, distortion is corrected for each split cell. be able to. Therefore, even when distortion due to variations in speed during imaging of an inspection target product or other causes occurs during imaging, a precise inspection image in which the distortion is corrected for every divided cell can be obtained. As a result, the accuracy of the inspection image is improved, and erroneous determination at the time of collation can be reduced as much as possible.

  In a preferred aspect, the interpolation calculation means sets a polygon that is divided into triangles having the reference point as a vertex, and performs a linear interpolation calculation on all divided cells other than the correction application cell among the divided cells. An assigning step for assigning to any polygon as a target; and a normal vector along the normal direction of the divided cell of the captured image with a scalar quantity corresponding to the deviation amount as a base point and the vertex of the polygon A normal vector setting step for each calculation, and a matrix calculation for calculating, for each component of the normal vector, the deviation amounts of all the divided cells assigned to the polygon based on the calculated end point of the normal vector. Steps are executed. In this aspect, a normal vector that reflects the deviation amount at the reference point is set in the polygon having the reference point of the correction application cell as a vertex, and all the divided cells are further based on the set normal vector. Since the amount of deviation is calculated by linear interpolation and a corrected inspection image is generated, the calculation load of the amount of deviation can be reduced, and correction processing can be executed more precisely and quickly.

  In a preferred aspect, there is provided an operation order storage means for storing the operation processing order of the divided cells assigned to the polygon for each polygon. In this aspect, since the order of the divided cells to be calculated can be preset for each polygon and the calculation process can be optimized, the calculation load can be further reduced and the process can be speeded up. .

  As described above, according to the present invention, even when distortion due to speed variation or other causes during imaging of an inspection target product occurs during imaging, the distortion is corrected for every divided cell. An image can be obtained. Therefore, the present invention has a remarkable effect that the image inspection of the inspection target product can be executed precisely even if the inspection target product is a flexible packaging container with high flexibility.

It is a perspective view which shows the tube as an example of the test object goods which concern on this invention. FIG. 2 is a schematic front view illustrating a configuration example of the tube inspection apparatus in FIG. 1 as an example of an embodiment of the present invention. 2 is a schematic plan view of the tube of FIG. It is a block diagram which shows the structural example of the inspection apparatus of FIG. It is a flowchart which shows the whole quality determination. It is a schematic diagram of the captured image which shows an example which performed the flowchart of FIG. It is a schematic diagram which shows the process of the captured image of FIG. FIG. 7 is a schematic diagram illustrating an enlarged process of the captured image in FIG. 6. FIG. 7 is a schematic diagram illustrating an enlarged process of the captured image in FIG. 6. It is an image figure of difference data which shows the result of having performed the flow chart of Drawing 5, (A) relates to this embodiment and (B) relates to a conventional product. It is a flowchart which shows the production | generation process of a reference | standard image. FIG. 12 is a schematic diagram of image data showing a process of processing a reference image based on the flowchart of FIG. 11. FIG. 12 is a schematic diagram of image data showing a process of processing a reference image based on the flowchart of FIG. 11. It is a schematic diagram of the image data explaining the processing process of FIG. FIG. 12 is a schematic diagram of image data showing a process of processing a reference image based on the flowchart of FIG. 11. FIG. 12 is a schematic diagram of image data showing a process of processing a reference image based on the flowchart of FIG. 11. FIG. 12 is a schematic diagram of image data showing a process of processing a reference image based on the flowchart of FIG. 11. 12 is a view table showing an example of data generated based on the flowchart of FIG. 11. It is a flowchart which shows the deviation amount calculation subroutine of all the cells of FIG. 20 is a flowchart showing a linear interpolation calculation subroutine of FIG. FIG. 21 is a flowchart showing a continuation of FIG. 20. It is a view table | surface which shows an example of the data regarding the process of a captured image. It is a perspective view which shows typically the example of execution of the flowchart of FIG. 19, FIG.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings.

  With reference to FIG. 1, the inspection apparatus according to the present embodiment uses a tube 1 as an example of an inspection target product illustrated in FIG. 1 as an inspection target. The tube 1 has a body portion 1a formed into a cylindrical shape, a flange portion 1b extending from the top of the body portion 1a, and a neck portion 1c projecting concentrically at the center of the flange portion 1b. It is a soft packaging material container.

  The main body 1a is formed of a highly flexible resin such as polyethylene or polypropylene. A plurality of patterns 1d are printed on the outer peripheral surface of the main body 1a. The inspection apparatus 10 according to the present embodiment inspects the quality of the pattern 1d of the main body 1a by image processing. Here, the “pattern” refers to a generic name such as a pattern, a character, a logo, and the like. At the inspection stage, the lower end of the main body 1a is opened. And when it determines with a good product after a test | inspection, the content is filled into the main-body part 1a, and the lower end of the main-body part 1a is sealed.

  Referring to FIG. 2, the inspection apparatus 10 includes an inspection table 11 that rotates about one horizontal axis. In the inspection table 11, along the tangential direction of the inspection table 11, a carry-in conveyor 12 for carrying in the tube 1 before the inspection, a non-defective product carry-out conveyor 14 for carrying out the tube 1 determined to be a good product after the inspection, and after the inspection A defective product carry-out conveyor 16 for carrying out the tube 1 determined to be defective is provided.

  A plurality of (for example, six) mandrels 11 a are equally arranged on the outer periphery of the inspection table 11. Each mandrel 11a protrudes in one direction (in the illustrated example, on the front side of the paper surface) in parallel with the rotation axis of the inspection table 11. The mandrel 11a is attached so as to be rotatable around its own central axis. The mandrel 11a is provided with a power input member (for example, a gear) 11b. When power is input to the power input member 11b, the mandrel 11a rotates about its own center line.

  On the outer periphery of the inspection table 11, a plurality of stations ST1 to ST6 are equally set corresponding to the number of mandrels 11a. The inspection table 11 rotates in one circumferential direction and intermittently conveys the mandrel 11a so that the mandrel 11a is temporarily stopped at each station ST1 to ST6.

  The downstream end of the carry-in conveyor 12 faces, for example, the station ST5. The upstream ends of the non-defective product delivery conveyor 14 and the defective product delivery conveyor 16 face stations ST4 and ST3 located downstream from the station ST5.

  In the station ST5, the tube 1 carried into the downstream end is attached to the mandrel 11a temporarily stopped in the station ST5 by a robot (not shown). In addition, at the stations ST4 and ST3, the tubes 1 determined to be good or bad are discharged to the upstream end of the conveyor 14 (16) corresponding to the determination result by a robot (not shown).

  An imaging station (for example, ST1) that executes an imaging process is set between the station ST5 of the carry-in conveyor 12 and the stations ST4 and ST3 of the carry-out conveyors 14 and 16.

  In the imaging station, a power output member 18 for outputting power to the power input member 11b of the mandrel 11a temporarily stopped is provided. The power output member 18 is, for example, a gear driven by a stepping motor, and rotates the mandrel 11a around the center line of the mandrel 11a while the mandrel 11a is stopped at the station ST1.

  In addition, the station ST1 receives a CCD camera 20 as an example of an imaging device, an LED line illumination device 21 that illuminates the tube 1 of the mandrel 11a, and light reflected from the tube 1 by being irradiated from the LED line illumination device 21. A mirror 22 leading to the CCD camera 20 is provided. The mandrel 11a to which the tube 1 is attached reaches the station ST1 via the station ST6. While the mandrel 11a stops at the station ST1 and receives the driving force of the power output member 18 and the mandrel 11a rotates once, the LED line illumination device 21 illuminates the tube 1 on the mandrel 11a, and the reflected light is CCD The camera 20 captures an image of the tube 1. Specifically, the control unit 100 described later generates an inspection image based on the captured image, and determines the quality of the inspection image.

  Here, with reference to FIG. 3 (A) (B), the problem when the tube 1 is mounted | worn with the mandrel 11a is demonstrated.

  As described above, the tube 1 to be inspected has the lower end of the main body 1a opened. Therefore, in the free state, as shown in FIG. 3 (A), the middle portion is bent more than the vicinity of the flange portion 1b having the diameter D1, and the major axis is in the elliptical shape having D2. This bending is somewhat mitigated by the attachment to the mandrel 11a, as shown in FIG. 3B, but the major axis D3 never becomes equal to the diameter D1. This is because a gap for inserting and removing the tube 1 from the mandrel 11a is required.

  For this reason, at the time of imaging at the station ST1, when imaging is performed by applying light in a direction parallel to the long axis D2, the peripheral speed of the outer peripheral surface of the tube 1 is increased, and the light shines in a direction intersecting the long axis D2. When the image is picked up, the peripheral speed of the outer peripheral surface of the tube 1 becomes slow. As a result, the image picked up by the CCD camera 20 includes distortion as the peripheral speed changes.

  The inspection apparatus 10 according to the present embodiment can remove this distortion by a simple method and correct it to a precise inspection image.

  Next, referring to FIG. 4, the inspection apparatus 10 includes an input device 30, a display device 40, and a storage device 50 in addition to the above-described inspection table 11, power output member 18, CCD camera 20, and the like. There is provided a control unit 100 for controlling the above.

  The input device 30 is embodied by a pointing device such as a keyboard or a mouse, for example, and is a unit for an operator to perform an input operation.

  The display device 40 is embodied by, for example, a liquid crystal display, and displays the processing state of the control unit 100 and an instruction screen of an interactive program to the operator using a GUI.

  The storage device 50 is embodied by an auxiliary storage device configured by a non-volatile memory, and stores various programs and OS executed by the control unit 100 and various data necessary for inspection processing.

  Next, the control unit 100 is a unit including a microprocessor, a memory, and the like. The distributed model processing module 101, the divided cell registration module 102, the correction application cell registration module 103, the deviation amount calculation module 104, the interpolation calculation module 105, An inspection image generation module 106 and a pass / fail determination module 107 are logically provided.

  Each process realized by these modules 101 to 107 will be described with reference to the flowcharts in FIG.

  With reference to FIG. 5, when the inspection apparatus 10 is operated, the control unit 100 reads the target reference image from the storage device 50 to the memory (step S <b> 1). Next, the mechanism shown in FIG. 2 is operated, and the tube 1 is mounted on the mandrel 11a of the inspection table 11 from the carry-in conveyor 12 and conveyed to the imaging station (station ST1). Here, the CCD is rotated while rotating the mandrel 11a. The camera 20 images the entire circumference of the tube body 1a over the entire circumference (step S2). The control unit 100 registers the captured image obtained by imaging in the storage device 50 for each product. FIG. 6 schematically shows an example of the captured image registered in step S2. As illustrated, in the captured image, the height direction of the tube 1 is set to X, and the upper side of the tube 1 is set to X +. Further, the circumferential direction of the tube 1 is set as Y in the captured image. As will be described later, an inspection image corrected for each divided cell CL divided by the matrix is generated from the captured image (see FIG. 7). In FIG. 7, D is a scratch to be detected. In the following description, when a specific divided cell CL is indicated, CL [i, j] is displayed, and the X coordinate and Y coordinate of the divided cell CL are i and j, respectively.

  Next, the control unit 100 executes a pattern matching process for each registered product (step S3). Based on this pattern matching processing, the control unit 100 detects the amount of deviation in the X direction and the Y direction for each correction application cell (the divided cell CL surrounded by a thick frame in FIG. 8) CLa set in the predetermined divided cell CL. dx and dy are respectively calculated (step S4), and the calculation result is registered in the storage device 50 (step S5). FIG. 8 shows an example in which the shift amounts dx and dy are calculated for the correction application cell CLa. The upper side of the divided cell CL surrounded by the thick frame indicates the coordinates [i, j] of the divided cell CL, and the lower side indicates the shift amounts dx and dy of the corresponding divided cell CL. The storage device 50 has an area for storing a calculated flag F2 indicating whether or not each divided cell CL has been processed (see FIG. 22), and the control unit 100 at this timing, shifts dx, dy The calculated flag F2 corresponding to the divided cell CL, that is, the correction application cell CLa is turned ON (step S6). Further, the storage device 50 has an area for storing a correction application flag F1 indicating that it is a correction application cell CLa, and the correction application flag F1 is also updated to ON.

  Next, the control unit 100 calculates the shift amounts dx and dy for each component for all the divided cells CL except the correction application cell (step S7), and based on the calculated shift amounts dx and dy, the captured image An inspection image with corrected distortion is generated (step S8). As a result, in the present embodiment, as shown in FIG. 9, it is possible to execute the shift correction based on the shift amounts dx and dy for every divided cell CL.

  Thereafter, the control unit 100 compares the reference image and the inspection image and calculates a difference (step S9). Specifically, the control unit 100 compares the density of each pixel of the inspection image with the corresponding pixel of the reference image, and determines a binary image in which a pixel having a large degree of deviation from the reference is 1 and the other is 0. And the area of the non-zero chunk is counted from the binary image.

  Next, an area is calculated for each counted chunk (step S10), and it is determined whether or not the area of each chunk is within an allowable range (step S11). If the area of all the blocks is within the allowable range, the control unit 100 performs a pass determination (step S12), puts the value of the non-defective product in the memory area for storing the pass / fail determination of the inspected product, and ends the pass / fail determination. On the other hand, if the area of any one lump D is outside the allowable range in the determination in step S11, the control unit 100 performs a failure determination and stores the value of the defective product in the memory area that stores the pass / fail determination of the inspection product. And the pass / fail determination is terminated (step S14).

  As shown in FIG. 10A, when the scratch D of the tube 1 is reflected, the scratch D is detected as a lump in step S10, and the control unit 100 performs the defect determination in the determination in step S11. . In addition, the binarized non-defective image is a black image without the block D.

  However, even if it is a non-defective product, the conventional processing method cannot completely remove the image distortion caused by the change in the peripheral speed as shown in FIG. As shown in the figure, the block De based on the distortion of the captured image is detected, which causes a misjudgment that determines that a non-defective product is defective.

  Next, a reference image generation example according to the present embodiment will be described with reference to FIG.

  First, the control unit 100 reads the image data that is the basis of the reference image (step S20). The image data is read by imaging a non-defective sample with an inspection apparatus. In this case, the captured image data may include distortion due to a change in peripheral speed. However, if the inspection image is also corrected in accordance with this distortion, the same effect as the comparison with an image without distortion is obtained as a result. It can be done.

  The captured image data is stored for each pixel as a development view of the main body 1a of the tube 1. In the present embodiment, at this stage, the control unit 100 registers the image data with the height of the main body 1a as the X direction and the circumferential direction as the Y direction in the development view. At this stage, for example, the image data shown in FIG. 12 is stored.

  Next, a pattern matching model (PM model) M as an example of a dispersion model is set in the image data (step S21). In this step, the operator displays a developed view as shown in FIG. 12 on the display device 40, and sets the characteristic part of the pattern as the PM model M. In this extraction process, it is preferable that the process proceeds interactively with an operator who operates the input device 30 so that the position of the PM model M is appropriately corrected and finally determined. Alternatively, it may be set automatically. In the present embodiment, when setting the PM model M, the operator includes that the PM model M includes at least one of the patterns 1d, and if the pattern continues for more than 1/4 turn in the Y direction, It is obliged to work as a constraint condition to arrange the PM model M in the middle. In this process, for example, the PM model M shown in FIG. 13 is set.

  Next, the control unit 100 sets the divided cell CL (step S22). As shown in FIG. 14, the divided cells CL are set in a matrix with predetermined pitches Scx and Scy determined in advance through experiments and simulations. Next, the PM model M is assigned to the divided cell CL. Since the PM model M itself is area data selected by reflecting the pattern of the image, for example, among the area data, the divided cell CL that occupies the largest area is selected, and the divided cell CL of the divided cell CL is selected. By selecting the center coordinates as the reference point P, an operation of assigning one PM model M to one divided cell CL is performed. The work of assigning the PM model M to the divided cell CL and setting the reference point P is preferably performed automatically by setting a predetermined algorithm, but is set by an interactive program for the operator. You may do it. The divided cell CL in which the reference point P is set is handled as the correction application cell CLa.

  Here, in the present embodiment, not only the divided cell CL in which the PM model M is set but also the other divided cells CL in which the PM model M is not set among the captured images to be inspected by the linear interpolation. dx and dy are obtained. In order to enable this linear interpolation, as shown in FIG. 15, a triangular polygon S having apexes of each reference point P is set, a divided cell CL is assigned to each polygon S, and a deviation dx, dy is calculated. By the way, in order to calculate the shift amounts dx and dy for all the divided cells CL, it is necessary that all the divided cells CL are included in any polygon S. However, depending on the design of the main body 1a of the tube 1, the reference point P that becomes the vertex of the polygon S may be insufficient. In that case, it becomes impossible to cover all the divided cells CL with the polygon S, and for example, the divided cells CL indicated by diagonal lines in FIG.

  Therefore, in step S22 of the present embodiment, the control unit 100 sets the first image data shown in the 0th to 9th rows in the Y coordinate of FIG. 15 as Y1, and the upstream and downstream sides of this image data Y1 in the Y direction. The extension data Y2 and Y3 are continuously arranged on the side to create an extension image for three rounds, and the image data is extended.

  Next, as shown in FIG. 16, the control unit 100 sets the divided cell CL including the reference point P as the correction application cell CLa (step S23). The reference point P of the correction application cell CLa is the basis for calculating the shift amounts dx and dy. By the way, it is preferable for reducing the error that the polygon S set at the time of linear interpolation is composed of short sides as much as possible. Further, it is necessary that the polygon S includes all the divided cells CL. In order to respond to such a request, in the present embodiment, a predetermined divided cell CL is forcibly set as the correction application cell CLa. Specifically, as indicated by the hatched portion in FIG. 16, if there is no reference point P on one side in both directions (both sides in the X direction) of the reference captured image, corrections on both sides in the X direction are performed. Divided cells (CL [0,5] and CL [9,2] in the example shown) adjacent to the application cell CLa (CL [1,5] and CL [8,2] in the example shown) Is forcibly set as the correction application cell CLa (step S24). Further, the four corners of the expanded captured image (in the illustrated example, CL [0, 19], CL [0, 20], CL [9, 19], CL [9, 20]) are selected (steps). S25). A reference point P is set at the center of these correction application cells CLa set forcibly. For the set reference point P, deviations dx and dy of the reference point P of the correction application cell CLa closest to the correction application cell CLa are set.

  Next, the control unit 100 generates a polygonal polygon having the reference point P of all the correction application cells CLa including the extension data Y2 and Y3 as vertices, and divides the polygon into triangular polygons S using a predetermined algorithm. (Step S26). As an algorithm for generating the polygon S, it is preferable to use a two-dimensional Delaunay division method. In the two-dimensional Delaunay division method, a polygon including each vertex is defined by using the reference point P of the correction application cell CLa on the plane as a vertex so that no correction application cell CLa is an aggregate of triangles on the surface. Divided into triangular polygons S. By setting such a polygon S, as shown in FIG. 17, a polygon S to which all the divided cells CL of the reference image can be assigned is set.

  After step S26 is completed, all the divided cells CL are assigned to the polygon S based on the data of the correction application cells CLa constituting the polygon S (step S27), and a triangular surface information table indicating the relation is created. (Step S28). FIG. 18 is a view table showing an example of the triangular surface information table. In the illustrated example, the divided cells CL of the reference image are listed for each product number of the tube 1 that is a product, and the polygon S is assigned to each divided cell CL. In the view table of FIG. 18, i and j indicate the coordinates (x, y) of the divided cell CL. F1 is a correction application flag indicating that it is a correction application cell. The initial value of the correction application flag F1 is Null. Id is an identification number that uniquely identifies the polygon S, and identifies which divided cell CL belongs to which polygon S. Furthermore, in the present embodiment, the order in which each divided cell CL is calculated is set for each polygon S. When one polygon S (for example, polygon S with Id = 1) is selected, the polygon S is assigned to that polygon S. The divided cells CL to which they belong are selected in a predetermined order (for example, CL [0,0] → CL [0,1] → CL [0,2]... dy can be calculated. Note that this order has a predetermined value (for example, −1) so that the last line can be identified for each polygon S, thereby speeding up the calculation.

  When step S28 ends, the control unit 100 graphically displays the triangular plane division status on the display device 30 (step S29), and ends the reference image generation process.

  Next, a subroutine (step S7 in FIG. 5) for calculating the shift amounts dx and dy of the divided cells will be described with reference to FIGS.

  In general, linear interpolation is a method of calculating the coordinates of a point between two known coordinates based on the coordinates of two known points. In the present embodiment, the reference points P of the correction application cells CLa constituting the vertices of the polygon S are used to calculate the shift amounts (and hence correction amounts) dx and dy of the divided cells CL by linear algebra. In order to execute such linear interpolation, a storage area as shown in FIG. 22 is set in the storage device 50 of the control unit 100 together with the data table of FIG. 18 described above, and processing of the deviation amounts dx and dy is performed. Can be managed.

  Referring to FIG. 22, an area for managing product numbers is set in the storage device 50 in order to manage the deviation amounts dx and dy for each product. The coordinates i and j of the divided cell CL are managed for each product number. In the divided cell CL, the calculated flag F2 is set together with the correction application flag F1. As will be described later, the calculation order of the divided cells CL changes at random according to the allocation status with respect to the polygon S. Therefore, the processing order is managed by setting the value of the calculated flag F2 for each divided cell CL. ing. The initial value of the calculated flag F2 is null, and the processed value is ON. Furthermore, an area for storing the values of the deviations dx and dy calculated for each divided cell CL is provided. The correction application cell CLa is registered in steps S3 to S5 in FIG. 5, and the forced correction application cell and the four corner correction application cells set in steps S24 and S25 in FIG. 11 are processed in steps S24 and S25 in FIG. The shift amounts dx and dy of the correction application cell CLa collated in (5) are registered as they are, and at this time, the calculated flag F2 is updated to ON (see step S6 in FIG. 5).

  Referring to FIGS. 19 and 23, first, in this subroutine, control unit 100 is configured to sequentially process shift amounts dx and dy of divided cells CL for each row indicated by the Y coordinate. In order to execute such processing, the control unit 100 initializes the X coordinate i and Y coordinate j of the divided cell CL to be processed (step S71). Next, the control unit 100 determines the presence / absence of a row having a number to be processed based on the Y coordinate j to be processed (step S72). If the value j of the row to be processed is equal to or greater than the set maximum value jmax, the control unit 100 returns to the main routine. When there is an unprocessed row, the control unit 100 determines whether or not the divided cell CL to be processed is in the row based on the value of the X coordinate i (step S73). If the column value i of the divided cell CL to be processed is less than the set maximum value imax, it is further determined whether or not the calculated flag F2 of the divided cell CL is Null ( Step S74) If it is Null, linear interpolation calculation is executed (step S75). By this linear interpolation calculation, the shift amounts dx and dy of the divided cell CL are precisely calculated. When the linear interpolation calculation in step S75 is completed, the control unit 100 increments the X coordinate i of the divided cell CL (step S76), proceeds to step S73, and repeats the above-described processing.

  When the X coordinate i is greater than or equal to the maximum value imax in step S73 (in the case of NO determination), that is, when all the divided cells CL in the column are calculated, the control unit 100 initializes the value of the X coordinate i. In addition, the value of the Y coordinate j is incremented (step S77), and then the process proceeds to step S72.

  If the calculated flag F2 is not Null in step S74, that is, if an ON value is entered (see FIG. 22), the process proceeds to step S76. Thereby, it is possible to speed up the processing by eliminating the redundant calculation.

  Next, details of the linear interpolation calculation (step S75) in FIG. 19 will be described with reference to FIG.

  Referring to FIG. 20, in the linear interpolation calculation (step S75), data of polygon S previously assigned to divided cell CL is referred to, and vertices A, B, and C of polygon S are set.

  For example, in the example of FIG. 23, the polygon S has CL [1,5], CL [4,6], and CL [4,3] assigned to A, B, and C, respectively. If the polygon S is determined, the correction application cell CLa constituting the polygon S can be specified. Therefore, in this example, the values of the shift amounts dx and dy set in the specified correction application cell CLa are used. Reference is made (see FIG. 22).

  Next, in this embodiment, in order to efficiently calculate the amount of deviation dx in the X direction and the amount of deviation dy in the Y direction, the X direction and the Y direction are combined using the variable a that identifies the X coordinate and the Y coordinate. To calculate. In the example of this embodiment, the variable a is first defined as x (step S742).

  Next, for each of the vertices A, B, and C of the polygon S, a normal vector in which the shift amount d (a) in the a direction (X direction after step S742 is executed) is taken in the normal direction (Z direction). nA (a), nB (a), nC (a) and the coordinates of the end point A (a) of one normal vector nA (a) are used as the start point, and the other normal vectors nB (a), nC (a ) And the unit vectors Uv (a) and Vv (a) toward the end points B (a) and C (a) and the determinant Det (a) are calculated (step S743).

  In this step, the shift amount d (a) is coordinate-transformed into a three-dimensional map having coordinates A (a), B (a), and C (a) as vertices, and a polygon S (a) is set. The polygon S (a) is inclined with respect to the polygon S with variation based on the distortion of the captured image.

を用いて、（１−１）式と（１−２）式の連立方程式から、ｓ、ｔを解くこととしている。 The unit vectors Uv (a) and Vv (a) are the center points T of the divided cells CL of the divided cells CL (for example, CL [3, 5] in FIG. 23) other than the correction application cell CLa assigned to the polygon S. To the coordinate T (a) intersecting with the polygon S (a) in the normal direction, and linearly interpolates the deviation d (a) at the center point T (= scalar amount of the normal vector nT (a)) It is for calculating with. That is, a vector Tv (a) starting from the coordinate A (a) and ending at the coordinate T (a) is
Tv (a) = s * Uv (a) + t * Vv (a) (1)
Therefore, rewriting equation (1) as a component relationship
xt = s * xu + t * xv (1-1)
yt = s * yu + t * yv (1-2)
zt = s * zu + t * zv (1-3)
Since (xu, yu), (xv, yv), and (xt, yt) in the equations (1-1) and (1-2) are known coordinates, the coefficient determinant Det is obtained by Kramel's formula. (A)

Is used to solve for s and t from the simultaneous equations (1-1) and (1-2).

  Next, the calculated values are stored for the a direction (step S744).

  Next, the control unit 100 determines whether or not the variable a is y, that is, whether or not the calculation has been completed for both the X direction and the Y direction (step S745). When the calculation is completed for both the X direction and the Y direction, the process proceeds to the next step. When the variable a is x, that is, when the calculation in the Y direction is not completed, the variable a is set to y. (Step S746), and the process proceeds to step S743.

  As shown in FIG. 20, with respect to one polygon S, A (a), B (a), C (a), Uv (a), Vv, which are common elements necessary for calculating the shift amounts dx, dy. By calculating (a) and Det (a) together, the shift amount calculation of the plurality of divided cells CL assigned to the polygon S can be speeded up.

  Next, when the processing in FIG. 20 is completed, as shown in FIG. 21, the control unit 100 again calculates the shift amount of the divided cell CL related to the polygon S in the X direction and the Y direction. a is defined as x (step S750). Next, the coordinates on the polygon S of CL [i, j] to be calculated are defined as T, and the coordinates indicating the shift amount are defined as T (a) (step S751). As a result, the vector Tv (a) having the coordinate A (a) as the start point and the coordinate T (a) as the end point is expressed by the unit vectors Uv (a) and Vv (a) as shown in the equation (1). The components xt, yt, and zt are also expressed by the equations (1-1) to (1-3).

を演算する（ステップＳ７５３）。 Next, the control unit 100 calculates xt and yt in the expressions (1-1) and (1-2) (step S752). In particular,
xt = (x coordinate of T (a)) − (x coordinate of A (a)) (3-1)
yt = (y coordinate of T (a)) − (y coordinate of A (a)) (3-2)
To obtain these values. If this calculation result is substituted into the equations (1-1) and (1-2), respectively, the equations (1-1) and (1-2) become simultaneous equations with s and t as variables. Accordingly, since s and t can be obtained by Kramel's formula, the control unit 100 uses the determinant Det (a) of the coefficient obtained by the equation (2) as a denominator.

Is calculated (step S753).

  Next, the control unit 100 substitutes the calculated values of s and t into the expression (1-3), and calculates the component zt in the normal direction of the vector T → T (a) (step S754). Here, zt in the expression (1-3) is a shift amount of the point T (a) in the normal direction with the point A (a) as the origin. Therefore, the control unit 100 adds the component in the normal direction from the point A to the point A (a), that is, the shift amount d (A (a)) of the point A to this zt, and the T at T (a) Is calculated and registered (step S755).

  Next, the control unit 100 determines whether or not a is y, that is, whether or not the calculation has been completed for both the X direction and the Y direction (step S756). If the calculation in the Y direction is not completed, a is changed to y (step S757), and the process proceeds to step S751.

  When the calculation is completed for both the X direction and the Y direction, the calculated flag F2 is turned ON for the divided cell CL (step S758). At this stage, the calculation of the shift amounts dx and dy related to the divided cell CL is completed.

  Next, the control unit 100 determines whether or not there is a divided cell CL for which calculation has not been completed for the polygon S that has been subjected to calculation (step S759). In step S750, the next divided cell CL is selected based on the calculation order Py (see FIG. 18), and the processing described above is repeated. On the other hand, when the calculation is completed for all the divided cells CL, the control unit 100 returns to the subroutine of FIG.

  As described above, in the present embodiment, the outer periphery of the highly flexible cylindrical tube 1 is imaged by being displaced relative to the CCD camera 20 and becomes a criterion for pass / fail of the tube 1 from the obtained captured image. In the inspection image generation method for generating an inspection image to be collated with a reference image, a divided cell setting step (step S22) for setting a divided cell CL obtained by dividing the reference image into a two-dimensional matrix, and among the divided cells CL, an inspection is performed. A correction application cell setting step (steps S23 to S25) for setting, as the correction application cell CLa, a coordinate including the coordinates representing the feature of the target pattern 1d as the reference point P, and the captured image and the reference image are collated by pattern matching. , A deviation amount detection step for detecting the deviation amounts dx and dy of the captured image corresponding to the reference point P included in the correction application cell CLa for each reference point P. Based on (step S4) and the shift amounts dx and dy at the reference point P detected in the shift amount detection step (step S4), the shift amounts dx and dy of all the divided cells CL other than the correction application cell CLa are linear. Based on the interpolation calculation step (step S7) calculated by interpolation, and the shift amounts dx and dy of all the divided cells CL including the correction application cell CLa, an inspection image for correcting the captured image for each divided cell CL is generated. And an image generation step (step S8).

  In the present embodiment, based on the precise shift amounts dx and dy obtained by pattern matching between the captured image and the reference image, the shift amounts dx and dy of the captured image are all divided cells CL by interpolation (step S7). Therefore, even if distortion due to speed variations during imaging of the tube 1 or other causes occurs at the time of imaging, a precise inspection image in which the distortion is corrected for every divided cell CL is obtained. Can do. As a result, the accuracy of the inspection image is improved, and erroneous determination at the time of collation can be reduced as much as possible.

  In this embodiment, the polygon setting step (step S26) for setting the polygon S divided into the triangles A, B, and C having the reference point P as the vertex, and the division cell CL other than the correction application cell CLa. An allocation step (step S27) in which all divided cells CL are assigned to any polygon S as a target of linear interpolation, and a captured image is divided for each vertex A, B, C of the polygon S corresponding to the shift amounts dx, dy. Normal vector setting step (step S743) for calculating vectors nA (a), nB (a), nC (a) along the normal direction Z of the cell CL, and the calculated normal vectors nA (a), nB (A) Based on the end points (A (a), B (a), C (a)) of nC (a), the shift amounts dx, dy of all the divided cells CL allocated to the polygon S are components. Calculate every That matrix and a calculating step (step S750~S760). For this reason, in this embodiment, in calculating linear interpolation, the end points (A () of normal vectors nA (a), nB (a), and nC (a) having the vertices A, B, and C of the triangular polygon as base points are calculated. Determinants can be set based on a), B (a), and C (a)), and the amount of deviation can be calculated by linear interpolation with a simple algebra.

  Further, in the present embodiment, a forcible setting step (steps S24 and S25) for forcibly setting a predetermined divided cell CL that contributes to generation of the polygon S (the divided cell in the hatched portion in FIG. 16) to the correction application cell CLa. I have. For this reason, in this embodiment, when setting the polygon S, it is possible to prevent the area of the polygon S from becoming excessively large or prevent the divided cells CL not included in the polygon S from being generated. Calculation errors of dx and dy can be suppressed.

  Further, in the present embodiment, an expansion step (step S22) in which the same reference image is continued on both sides of the reference image, and a divided cell CL corresponding to the correction application cell CLa among the reference images continued on both sides. And registering the divided cells CL at the corners of the reference image continuous on both sides as correction application cells including the reference point P (steps S24 and S25). Therefore, in this embodiment, all the divided cells CL can be assigned to the polygon S, and the deviation amounts dx and dy for each divided cell CL can be reliably calculated by linear interpolation.

  In this embodiment, the polygon setting step (step S26) includes a step of dividing the polygonal polygon into triangular polygons S by the two-dimensional Delaunay division method (see FIG. 17). For this reason, in this embodiment, suitable three-dimensional mapping which can cover each correction application cell as a vertex can be performed.

  Further, in the present embodiment, the matrix calculation step uses the end point A (a) of one normal vector nA (a) calculated with the vertex A of the polygon S as a base point as a starting point, and configures the polygon S. Two unit vectors Uv (a) and Vv (a) toward the end points B (a) and C (a) of the normal vectors nB (a) and nC (a), and the unit vectors Uv (a) and Vv A coefficient determinant (equation (2)) based on the component vector of (a) is calculated in advance, and assigned to the polygon S based on these unit vectors Uv (a), Vv (a) and the determinant. The step includes sequentially calculating the shift amounts dx and dy for each divided cell CL. For this reason, in the present embodiment, the calculation of the divided cells CL is sequentially executed in a state in which the unit vectors Uv (a), Vv (a) and determinants necessary for calculating the shift amounts dx, dy are calculated in advance. Therefore, it is possible to eliminate duplication processing and improve the efficiency of calculation.

  According to another aspect of the present invention, in the image inspection method using the above-described inspection image generation method, the inspection image generated in the inspection image generation step (step S8) is compared with the reference image, and the inspection image This is an image inspection method including a pass / fail determination step (steps S11 to S14) for determining pass / fail.

  Another embodiment of the present invention includes a mandrel 11a that supports a highly flexible cylindrical tube 1 rotatably in one direction, a CCD camera 20 that captures an outer periphery of the tube 1 supported on the mandrel 11a, and a CCD. Reference image storage means for storing a reference image serving as a reference for pass / fail of the tube 1 imaged by the camera 20, and reference point storage means for storing, as a reference point P, coordinates representing the characteristics of the pattern to be inspected among the reference images. And a divided cell information storage means for storing information on the divided cells CL obtained by dividing the reference image into a two-dimensional matrix, and a correction application as a means for discriminating a cell including the reference point P from the divided cells CL from the correction application cell CLa. The storage device 50 embodies correction application cell storage means for storing the flag F1. In addition, the control unit 100 collates the captured image obtained by the CCD camera 20 with the reference image by pattern matching, and determines the shift amounts dx and dy of the captured image corresponding to the correction application cell CLa for each reference point P. A deviation amount calculation module 104 as a deviation amount detection means to be detected, and the deviation amounts dx and dy of all the divided cells CL based on the deviation amounts dx and dy at the reference point P detected by the deviation amount calculation module 104 are linearly interpolated. Based on the interpolation calculation module 105 as the interpolation calculation means for calculating, and the shift amounts dx and dy of the respective divided cells CL, the inspection image generation as the generation means is obtained by correcting the captured image for every divided cell CL. Judgment of determining whether the tube 1 is good by comparing the module 106 with the inspection image corrected by the inspection image generation module 106 and the reference image The acceptance judgment module 107 as a step comprises logically. Therefore, in this embodiment, based on the shift amounts dx and dy obtained by pattern matching, the shift amounts dx and dy of all the divided cells CL of the captured image are calculated by linear interpolation, and the inspection image is corrected for distortion. Therefore, distortion can be corrected for each divided cell. Therefore, even if distortion due to speed variation during imaging of the tube 1 or other causes occurs during imaging, a precise inspection image in which the distortion is corrected for each divided cell CL can be obtained. As a result, the accuracy of the inspection image is improved, and erroneous determination at the time of collation can be reduced as much as possible.

  In the present embodiment, the interpolation calculation module 105 includes a polygon setting step (step S26) for setting a polygon S divided into triangles having the reference point P as vertices A, B, and C, and among the divided cells CL, An assigning step (step S27) in which all divided cells CL other than the correction application cell are assigned to any polygon S as a target of linear interpolation, and the shift amounts dx and dy are divided into the divided cells for each vertex A, B, and C of the polygon S. A normal vector setting step (step S741) for calculating normal vectors nA (a), nB (a), nC (a) along the normal direction Z of CL, and the calculated normal vector nA (a), Based on the end points (A (a), B (a), C (a)) of nB (a) and nC (a), the shift amounts dx and dy of all the divided cells CL assigned to the polygon S are calculated. component And it executes a matrix arithmetic operation step of calculating (step S75) to. Therefore, in the present embodiment, normal vectors nA (a), nB (a), which reflect the deviations dx and dy at the reference point P on the polygon S having the reference point P of the correction application cell CLa as the vertex, nC (a) is set, and based on the set normal vectors nA (a), nB (a), and nC (a), the shift amounts dx and dy of all the divided cells CL are calculated by linear interpolation. Thus, a corrected inspection image is generated, so that the calculation load of the shift amounts dx and dy can be reduced, and the correction process can be executed more precisely and quickly.

  Further, in the present embodiment, there is provided an operation order storage means for storing an attribute (order of FIG. 18) for specifying the operation processing order of the divided cells CL assigned to the polygon S for each polygon S. For this reason, in this embodiment, the order of the divided cells CL to be calculated can be preset for each polygon S to optimize the calculation process, so that the calculation load can be further reduced and the process can be performed quickly. Can be

  The above-described embodiment is merely a preferred specific example of the present invention, and the present invention is not limited to the above-described embodiment. For example, in the calculation step S7 of the shift amounts dx and dy of all cells, the divided cells CL may be advanced by increasing the rows for each column instead of increasing the columns for each row and moving the divided cells CL.

  It goes without saying that various modifications can be made within the scope of the claims of the present invention.

DESCRIPTION OF SYMBOLS 1 Tube 1a Tube main-body part 1d Pattern 10 Inspection apparatus 11 Inspection table 11a Mandrel 20 CCD camera (an example of imaging device)
DESCRIPTION OF SYMBOLS 50 Memory | storage device 100 Control unit 101 Distributed model processing module 102 Divided cell registration module 103 Correction application cell registration module 104 Deviation amount calculation module 105 Interpolation calculation module 106 Inspection image generation module 107 Pass / fail judgment module nA (a), nB (a), nC (a) Normal vector A (a), B (a), C (a) End point of normal vector M PM model (an example of a distributed model)
Py calculation order (example of attribute)
S polygon

Claims (10)

In the inspection image generation method of generating an inspection image that is imaged by relatively displacing the inspection target product with the imaging device, and generating an inspection image that is collated with a reference image that is a reference for pass / fail of the inspection target product.
A divided cell setting step for setting a divided cell obtained by dividing the reference image into a two-dimensional matrix;
A correction application cell setting step for setting, as a correction application cell, one of the divided cells that includes coordinates representing the characteristics of the pattern to be inspected as a reference point;
A deviation amount detection step of collating the captured image with the reference image by pattern matching and detecting a deviation amount of the captured image corresponding to the correction application cell for each reference point;
Based on the amount of deviation at the reference point detected in the amount of deviation detection step, an interpolation calculation step for calculating the amount of deviation of all divided cells other than the correction application cell in the captured image by linear interpolation;
An inspection image generation method comprising: an inspection image generation step for generating an inspection image in which the captured image is corrected for each divided cell based on a shift amount of all the divided cells including the correction application cell. . The method for generating an inspection image according to claim 1,
A polygon setting step for setting a polygon divided into triangles having the reference point as a vertex;
Assigning all the divided cells other than the correction application cell among the divided cells to any polygon as a calculation target of linear interpolation;
A normal vector setting step for calculating, for each vertex of the polygon, a normal vector along the normal direction of the divided cell of the captured image with a scalar amount corresponding to the deviation amount as a base point; and
A matrix calculation step of calculating, for each component of the normal vector, the shift amounts of all the divided cells allocated to the polygon based on the end point of the calculated normal vector. Method for generating inspection images. The method for generating an inspection image according to claim 2,
An inspection image generation method, comprising: a forcible setting step for forcibly setting the divided cells contributing to generation of the polygon as correction application cells. The method of generating an inspection image according to claim 2 or 3,
An expansion step of continuing the same reference image on both sides of the reference image;
Registering the divided cells corresponding to the correction application cells and the divided cells at the corners of the reference image continuous on both sides of the reference images continuous on both sides as correction application cells. A method for generating an inspection image characterized by the above. The method for generating an inspection image according to any one of claims 2 to 4,
The polygon setting step includes a step of dividing a polygonal polygon into triangular polygons by a two-dimensional Delaunay division method. In the production | generation method of the test | inspection image in any one of Claim 2 to 5,
In the matrix calculation step, the end point of one normal vector calculated with the vertex of the polygon as a base point is set as a start point, and the end point of another normal vector calculated with another vertex constituting the polygon as a base point is set. Two unit vectors heading and determinants of coefficients based on the component vectors of each unit vector are calculated in advance, and based on these unit vectors and determinants, the shift amount for each divided cell assigned to the polygon is sequentially A method for generating an inspection image, comprising a step of calculating. In the image inspection method using the inspection image generation method according to claim 1,
An image inspection method comprising: a pass / fail judgment step of collating the inspection image generated by the inspection image generation method with the reference image and determining the pass / fail of the inspection image. An imaging device for imaging an inspection site of a product to be inspected;
Reference image storage means for storing a reference image which is a reference for pass / fail of the inspection target product;
Reference point storage means for storing coordinates representing the characteristics of the pattern to be inspected among the reference images as reference points;
Divided cell information storage means for storing information on divided cells obtained by dividing the reference image into a two-dimensional matrix;
A correction application cell storage unit that stores, as the correction application cell, the cell including the reference point among the divided cells;
A deviation amount detecting means for collating a captured image captured by the imaging device with the reference image by pattern matching, and detecting a deviation amount of the captured image corresponding to the correction application cell for each reference point;
Based on the deviation amount at the reference point detected in the deviation amount detection step, interpolation calculation means for calculating deviation amounts of all the divided cells other than the correction application cell in the captured image by linear interpolation;
Based on the amount of shift of each divided cell, inspection image generation means for generating an inspection image in which the captured image is corrected for every divided cell;
An appearance inspection apparatus comprising: a determination unit that compares the inspection image with the reference image to determine whether the inspection target product is good or bad. The appearance inspection apparatus according to claim 8,
The interpolation calculation means includes
A polygon setting step for setting a polygon divided into triangles having the reference point as a vertex;
Assigning all the divided cells other than the correction application cell among the divided cells to any polygon as a calculation target of linear interpolation;
A normal vector setting step for calculating, for each vertex of the polygon, a normal vector along the normal direction of the divided cell of the captured image with a scalar amount corresponding to the deviation amount as a base point; and
A matrix calculation step of calculating, for each component of the normal vector, a shift amount of all the divided cells assigned to the polygon based on the calculated end point of the normal vector. Appearance inspection device. The appearance inspection apparatus according to claim 9,
An appearance inspection apparatus, comprising: an operation order storage unit that stores the operation processing order of the divided cells assigned to the polygon for each polygon.

Images