Image generator, pre-trained model generator, and x-ray inspection device

Abstract

To provide an image generator, a pre-trained model generator, and an X-ray inspection device which can easily create a large amount of teacher data necessary for machine learning.SOLUTION: A server 50 includes: an image acquisition unit 53 for acquiring a non-defective image G1 on the basis of an X-ray which has passed through an item A without a defect F; and an image generation unit 54 for changing the pixel value of at least a part of a region which corresponds to the item A in the non-defective image G1 acquired by the image acquisition unit 53 and generating a defective image G2 including a virtual defect F in the item A. The image generation unit 54 generates a defective item image G2 including a virtual defect F in the inside of the item A, the attenuation rate of an X-ray in the virtual defect F being different from that of an X-ray in the item A.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an image generation device, a learned model generation device, and an X-ray inspection device.

Background Art

[0002] An X-ray inspection device includes a conveyance unit that conveys an article, an X-ray irradiation unit that irradiates the article conveyed by the conveyance unit with X-rays, an X-ray detection unit that detects the X-rays transmitted through the article, and an inspection unit that generates an X-ray inspection image from the X-rays detected by the X-ray detection unit and inspects the article based on the X-ray inspection image (see, for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In an X-ray inspection device, inspections are required to determine foreign substances such as resin contained inside an article as an abnormality, or to determine cavities (voids) formed inside the article as an abnormality. Foreign substances and cavities inside the article are difficult to clearly show edges in an X-ray inspection image. Therefore, it is difficult to set parameters for inspecting these abnormalities by image processing (energy analysis) in the X-ray inspection device.

[0005] Therefore, in an X-ray inspection apparatus, a pre-trained model (AI) can be used to distinguish between foreign objects and cavities inside an article and perform inspections. In order to improve the accuracy of inspections using the pre-trained model, it is necessary to perform machine learning using a large amount (an enormous amount) of teacher data (images) in the pre-trained model. In order to generate a large amount of teacher data, it is necessary to create samples in which the articles contain defects, but it is not easy to create a large number of these samples. Therefore, it has been difficult to obtain the large amount of teacher data required for machine learning.

[0006] One aspect of the present invention aims to provide an image generation apparatus, a pre-trained model generation apparatus, and an X-ray inspection apparatus that can easily create a large amount of teacher data required for machine learning.

Means for Solving the Problem

[0007] (1) An image generation apparatus according to one aspect of the present invention includes an image acquisition unit that acquires a first image based on X-rays that have passed through an article that does not contain a defect, and an image generation unit that changes the pixel values of at least a part of the region corresponding to the article in the first image acquired by the image acquisition unit to generate a second image in which a virtual defect is included in the article. The image generation unit generates a second image in which a virtual defect is included inside the article and in which the X-ray attenuation rate in the virtual defect is different from the X-ray attenuation rate in the article.

[0008] In an image generation device according to an aspect of the present invention, in a first image acquired by an image acquisition unit, a pixel value of at least a part of a region corresponding to an article is changed to generate a second image in which a virtual defect is included in the article. The first image is a good product image when the article is a good product, and the second image is a defective product image when the article is defective. The image generation unit generates a second image in which a virtual defect is included inside the article and the X-ray attenuation rate in the virtual defect is different from the X-ray attenuation rate in the article. The defect included inside the article has an X-ray attenuation rate different from that of the article. Thus, the image generation device can generate a second image with a different X-ray attenuation rate without creating a sample in which a defect is included inside the article. Therefore, the image generation device can easily create a large amount of teacher data required for machine learning.

[0009] (2) In the image generation device of (1) above, the image generation unit may generate a second image such that the X-ray attenuation rate in the virtual defect is lower than the X-ray attenuation rate in the article. Defects such as resin have a lower X-ray attenuation rate than the article. Therefore, the image generation device can generate a second image in which a virtual defect such as resin is included inside the article.

[0010] (3) In the image generation device of (1) or (2) above, the image generation unit may change a plurality of adjacent pixel values in at least a part of the region corresponding to the article. In this configuration, the pixel values of a continuous region are changed. Therefore, a second image in which a virtual defect such as resin is included inside the article can be appropriately generated.

[0011] (4) A learned model generation device according to an aspect of the present invention includes any one of the image generation devices of (1) to (3) above, and performs machine learning using the second image generated by the image generation unit in the image generation device as teacher data to generate a learned model for determining whether or not a defect is included in an article.

[0012] The trained model generation device according to one aspect of the present invention includes the above-described image generation device. Therefore, in the trained model device, a large amount of teaching data necessary for machine learning can be easily created. Further, in the trained model, since machine learning can be performed using a large amount of teaching data, the accuracy of the trained model can be improved.

[0013] (5) The trained model generation device according to one aspect of the present invention includes a learning unit that performs machine learning using the second image generated by the image generation unit in any one of the image generation devices (1) to (3) as teaching data, and generates a trained model for determining whether or not a defect is included in an article.

[0014] In the trained model generation device according to one aspect of the present invention, machine learning is performed using the teaching data generated by the image generation device. In the image generation device, a large amount of teaching data necessary for machine learning can be easily created. As a result, in the trained model generation device, since machine learning can be performed using a large amount of teaching data, the accuracy of the trained model can be improved.

[0015] (6) The X-ray inspection device according to one aspect of the present invention includes a conveyance unit that conveys an article, an X-ray irradiation unit that irradiates the article conveyed by the conveyance unit with X-rays, an X-ray detection unit that detects the X-rays that have passed through the article after being irradiated by the X-ray irradiation unit, and an inspection unit that generates an X-ray inspection image based on the detection result of the X-ray detection unit and inspects the article based on the X-ray inspection image. The inspection unit determines whether or not a defect is included in the article using the trained model generated by the trained model generation device in (4) above.

[0016] In the X-ray inspection device according to one aspect of the present invention, inspection is performed using the trained model generated by the above-described trained model generation device. Since the trained model generation device can perform machine learning using a large amount of teaching data generated by the image generation device, the accuracy of the trained model can be improved. In the X-ray inspection device, since this trained model is used for inspection, the inspection accuracy can be improved.

[0017] (7) The X-ray inspection apparatus according to one aspect of the present invention includes a conveyance unit that conveys an article, an X-ray irradiation unit that irradiates the article conveyed by the conveyance unit with X-rays, an X-ray detection unit that detects the X-rays irradiated by the X-ray irradiation unit and transmitted through the article, and an inspection unit that generates an X-ray inspection image based on the detection result of the X-ray detection unit and inspects the article based on the X-ray inspection image. The inspection unit determines whether or not the article contains a defect using the learned model generated by the learned model generation apparatus in (5) above.

[0018] In the X-ray inspection apparatus according to one aspect of the present invention, inspection is performed using the learned model generated by the above-described learned model generation apparatus. Since the learned model generation apparatus can perform machine learning using a large amount of teacher data generated by the image generation apparatus, the accuracy of the learned model can be improved. In the X-ray inspection apparatus, since this learned model is used for inspection, the inspection accuracy can be improved.

[0019] (8) The X-ray inspection apparatus according to one aspect of the present invention includes any one of the image generation apparatuses in (1) to (3) above, a learning unit that performs machine learning using the second image generated by the image generation unit in the image generation apparatus as teacher data and generates a learned model for determining whether or not the article contains a defect, a conveyance unit that conveys the article, an X-ray irradiation unit that irradiates the article conveyed by the conveyance unit with X-rays, an X-ray detection unit that detects the X-rays irradiated by the X-ray irradiation unit and transmitted through the article, and an inspection unit that generates an X-ray inspection image based on the detection result of the X-ray detection unit and inspects the article based on the X-ray inspection image. The inspection unit determines whether or not the article contains the defect using the learned model generated by the learning unit.

[0020] In the X-ray inspection apparatus according to one aspect of the present invention, since machine learning can be performed using a large amount of teacher data generated by the image generation apparatus, the accuracy of the learned model generated in the learning unit can be improved. As a result, in the X-ray inspection apparatus, the inspection accuracy can be improved.

Advantages of the Invention

[0021] According to one aspect of the present invention, a large amount of teaching data required for machine learning can be easily created.

Brief Description of Drawings

[0022]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Embodiments for Carrying Out the Invention

[0023] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same or corresponding elements are denoted by the same reference numerals, and duplicate descriptions are omitted.

[0024] As shown in FIG. 1, the X-ray inspection system 1 includes an X-ray inspection device 10, a sorting device 30, and a server (image generation device, learned model generation device) 50. The X-ray inspection system 1 inspects the article A using a learned model generated by machine learning. In the present embodiment, for example, the X-ray inspection system 1 inspects for defects existing inside the article A in an article A such as a soft candy. The defect is a foreign substance (for example, resin or the like) or a cavity inside the article A, but depending on the article A, a cavity may not be regarded as a defect.

[0025] The X-ray inspection device 10 includes a device main body 11, support legs 12, a shield box 13, a conveyance unit 14, an X-ray irradiation unit 15, an X-ray detection unit 16, a display operation unit 17, and a control unit (inspection unit) 18.

[0026] The X-ray inspection apparatus 10 generates an X-ray inspection image of the article A while conveying the article A, and inspects (inspects for defective inclusion) the article A based on the X-ray inspection image. The article A before inspection is carried into the X-ray inspection apparatus 10 by the carry-in conveyor 19. The article A after inspection is carried out to the conveyor 31. The article A determined to be a defective product (having an abnormality) by the X-ray inspection apparatus 10 is sorted out of the production line by the sorting device 30 arranged on the conveyor 31. The article A determined to be a non-defective product by the X-ray inspection apparatus 10 passes through the sorting device 30 as it is.

[0027] The apparatus main body 11 houses the control unit 18 and the like. The support legs 12 support the apparatus main body 11. The shield box 13 is provided on the apparatus main body 11. The shield box 13 prevents leakage of X-rays to the outside. Inside the shield box 13, an inspection area R where the article A is inspected by X-rays is provided. The shield box 13 is formed with a carry-in port 13a and a carry-out port 13b. The article A before inspection is carried into the inspection area R from the carry-in conveyor 19 through the carry-in port 13a. The article A after inspection is carried out from the inspection area R to the conveyor 31 of the sorting device 30 through the carry-out port 13b. X-ray shielding curtains (not shown) for preventing leakage of X-rays are provided at each of the carry-in port 13a and the carry-out port 13b.

[0028] The conveyance unit 14 is arranged inside the shield box 13. The conveyance unit 14 conveys the article A along the conveyance direction D from the carry-in port 13a through the inspection area R to the carry-out port 13b. The conveyance unit 14 is, for example, a belt conveyor stretched between the carry-in port 13a and the carry-out port 13b.

[0029] The X-ray irradiation unit 15 is arranged inside the shield box 13. The X-ray irradiation unit 15 irradiates the article A conveyed by the conveyance unit 14 with X-rays. The X-ray irradiation unit 15 has, for example, an X-ray tube that emits X-rays and a collimator that spreads the X-rays emitted from the X-ray tube in a fan shape in a plane perpendicular to the conveyance direction D.

[0030] The X-ray detection unit 16 is disposed within the shield box 13. The X-ray detection unit 16 is a line sensor composed of X-ray detection elements arranged in a one-dimensional manner along a horizontal direction perpendicular to the conveyance direction D. The X-ray detection unit 16 detects the X-rays that have passed through the article A and the conveyor belt of the conveyance unit 14.

[0031] The display operation unit 17 is provided on the apparatus main body 11. The display operation unit 17 displays various information and accepts input of various conditions. The display operation unit 17 is, for example, a liquid crystal display and displays an operation screen as a touch panel. In this case, the operator can input various conditions via the display operation unit 17.

[0032] The control unit 18 is disposed within the apparatus main body 11. The control unit 18 controls the operations of each part of the X-ray inspection apparatus 10. The control unit 18 is composed of a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), etc. The X-ray detection result of the X-ray detection unit 16 is input to the control unit 18. The control unit 18 creates an X-ray inspection image based on the X-ray detection result.

[0033] The control unit 18 inspects the article A based on the X-ray inspection image. The control unit 18 determines whether or not the article A contains a defect F using a learned model (described later) generated by the server 50. The control unit 18 outputs various signals based on the inspection information indicating the inspection result. The control unit 18 outputs a display signal for causing the display operation unit 17 to display the inspection result to the display operation unit 17 based on the inspection information. When the article A contains a defect F in the inspection information, the control unit 18 outputs a sorting signal for instructing the sorting of the article A to the sorting apparatus 30.

[0034] The sorting device 30 is provided on the downstream side of the X-ray inspection device 10. The sorting device 30 is provided on the conveyor 31. The sorting device 30 sorts the article A based on the sorting signal output from the X-ray inspection device 10. The sorting device 30 has a photoelectric sensor 32 and an arm 33.

[0035] The photoelectric sensor 32 is a sensor that detects the passage of the article A. The photoelectric sensor 32 is installed upstream of the arm 33 and detects the loading of the article A into the sorting device 30.

[0036] The arm 33 swings at the tip around the base end as the axis by a driving force such as a motor. The arm 33 pushes out the article A to one side in the width direction of the conveyor 31 and sorts the article A outside the production line.

[0037] As shown in FIG. 1, the server 50 is a device that generates a learned model by machine learning and performs processing by the learned model. The server 50 functions as an image generation device that generates teacher data (teacher images) used for generating the learned model. The server 50 functions as a learned model generation device that performs machine learning using the teacher data to generate a learned model. The server 50 is composed of a CPU, a ROM, a RAM, etc.

[0038] The server 50 is managed by, for example, the user of the X-ray inspection device 10. As shown in FIG. 1, the X-ray inspection device 10 and the server 50 are communicably connected by a wired or wireless network N such as the Internet or a telephone network, and can transmit and receive information to and from each other.

[0039] The server 50 includes a communication unit 51, a storage unit 52, an image acquisition unit 53, an image generation unit 54, and a learning unit 55.

[0040] The communication unit 51 communicates with the X-ray inspection apparatus 10. The communication unit 51 receives the image information transmitted from the X-ray inspection apparatus 10 and outputs it to the image acquisition unit 53. The communication unit 51 transmits the learned model to the X-ray inspection apparatus 10 in response to a request from the X-ray inspection apparatus 10.

[0041] The storage unit 52 stores various types of information. The storage unit 52 stores the information output from the communication unit 51. The storage unit 52 stores the defective product image G2 (described later) generated by the image generation unit 54. The storage unit 52 stores the learned model generated in the learning unit 55.

[0042] The image acquisition unit 53 acquires a non-defective product image (first image) G1 (see FIG. 2) based on the X-ray that has passed through the article A that does not contain defects. The image acquisition unit 53 acquires the X-ray inspection image stored in the storage unit 52 as the non-defective product image G1. As shown in FIG. 2, in the non-defective product image G1, for example, a plurality of articles A (six in FIG. 2) are displayed. In the example shown in FIG. 2, none of the articles A contain defects. The image acquisition unit 53 outputs the acquired non-defective product image G1 to the image generation unit 54.

[0043] As shown in FIG. 1, the image generation unit 54 changes at least a part of the pixel values (luminance values) in the region corresponding to the article A in the non-defective product image G1 acquired by the image acquisition unit 53, and generates a defective product image (second image) G2 in which a virtual defect F (see FIG. 3) is included in the article A. The image generation unit 54 generates a defective product image G2 in which a virtual defect F is included inside the article A. The inside of the article A is the portion on the path through which the X-ray passes between the front surface (the surface facing upward when being conveyed) and the back surface (the surface facing the conveying surface when being conveyed) of the article A.

[0044] The image generation unit 54 generates a defective product image G2 in which the X-ray attenuation rate in the virtual defect F is different from the X-ray attenuation rate in the article A. In the present embodiment, the image generation unit 54 generates the defective product image G2 such that the X-ray attenuation rate in the virtual defect F is lower than the X-ray attenuation rate in the article A. The image generation unit 54 changes at least a part of the pixel values adjacent to each other in a region corresponding to the article A. Specifically, for example, the image generation unit 54 changes the pixel values such that the pixel values in the region of the defect F (a region composed of a plurality of adjacent pixel values) are higher than the pixel values in the region corresponding to the article A. The image generation unit 54 changes the pixel values such that the pixel values in the region of the defect F are higher than the pixel values in the region corresponding to the article A by, for example, a value in the range of "10 to 20". The image generation unit 54 may change each of the plurality of pixel values by the same value, or may change the values such that the pixel values after the change are the same. Further, the image generation unit 54 may change the values of the pixel values to be changed according to the shape of the defect F.

[0045] The image generation unit 54 generates the specified number of defective product images G2 by the user. The image generation unit 54 generates the defective product image G2 by randomly including the defect F in the article A based on the non-defective product image G1. The image generation unit 54 randomly sets the number of defects F in the defective product image G2, the number of defects F in the article A, the position of the defect F in the article A, the size of the defect F, the pixel value of the defect F, etc.

[0046] The image generation unit 54 generates, for example, a defective product image G2 in which one article A includes the defect F, or generates a defective product image G2 in which each of a plurality of articles A includes the defect F. The image generation unit 54 generates a defective product image G2 in which one or a plurality of defects F are included in one article A. When the image generation unit 54 includes a plurality of defects F in one article A, the image generation unit 54 sets the positions of the defects F so that the plurality of defects F do not overlap. The image generation unit 54 generates a defective product image G2 including a defect F having a shape such as a circular shape, an elliptical shape, a triangular shape, a rectangular shape, a polygonal shape, etc. The shape of the defect F can be set to a shape corresponding to the shape of an actual defect. The image generation unit 54 stores the generated defective product image G2 in the storage unit 52.

[0047] As shown in FIG. 3, a plurality of articles A (six in FIG. 3) are displayed in the defective article image G2 generated by the image generation unit 54. In the example shown in FIG. 3, the defective article image G2 includes a defect F in one article A. The defect F has, for example, a triangular shape. In the example shown in FIG. 4, the defective article image G2 includes a defect F in each of two articles A. One article A includes a defect F having a square shape, and the other article A includes a defect F having a circular shape.

[0048] As shown in FIG. 1, the learning unit 55 acquires teacher data (learning data) used for machine learning, and performs machine learning using the acquired teacher data to generate a learned model. The learning unit 55 performs machine learning using the defective article image G2 generated by the image generation unit 54 as teacher data, and generates a learned model for determining whether or not the article A includes a defect F.

[0049] The learning unit 55 uses each pixel value of the teacher image as an input value to the neural network, and performs machine learning with the processing information corresponding to the teacher image as the output value of the neural network to generate the neural network. When using the pixel value as the input value, it is set as the input value of the neuron associated with each pixel (the position of the pixel on the image). The machine learning itself can be performed in the same manner as a conventional machine learning algorithm. The learning unit 55 stores the generated learned model in the storage unit 52.

[0050] Subsequently, the operation (inspection method) of the X-ray inspection system 1 will be described with reference to FIG. 5. As shown in FIG. 5, the X-ray inspection apparatus 10 creates an X-ray inspection image from the detection result of the X-ray detection unit 16 (step S01). The X-ray inspection apparatus 10 transmits the X-ray inspection image to the server 50 (step S02). Subsequently, the server 50 acquires a non-defective article image G1 based on the X-ray inspection image (step S03), and generates a defective article image G2 based on the non-defective article image G1 (step S04).

[0051] The server 50 performs machine learning based on the defective product image G2 and generates a learned model (step S05). The server 50 transmits the learned model to the X-ray inspection apparatus 10 (step S06).

[0052] The X-ray inspection apparatus 10 inspects whether or not the article A contains a defect F based on the learned model (step S07). When the X-ray inspection apparatus 10 determines that the article A contains a defect F (step S08: YES), it transmits a sorting signal to the sorting apparatus 30, and the sorting apparatus 30 sorts the article A (step S09). When the X-ray inspection apparatus 10 determines that the article A does not contain a defect F (step S08: NO), the process ends.

[0053] As described above, the server 50 of the X-ray inspection system 1 according to the present embodiment includes an image generation unit 54 that changes at least a part of the pixel values in the region corresponding to the article A in the non-defective product image G1 acquired by the image acquisition unit 53, and generates a defective product image G2 in which a virtual defect F is included in the article A. The image generation unit 54 generates a defective product image G2 in which a virtual defect F is included inside the article A and the X-ray attenuation rate in the virtual defect F is different from the X-ray attenuation rate in the article A. The defect F included inside the article A has a different X-ray attenuation rate from that of the article A. Thus, the server 50 can generate a defective product image G2 with different X-ray attenuation rates without creating a sample in which the defect F is included inside the article A. Therefore, the server 50 can easily create a large amount of teaching data required for machine learning.

[0054] In the X-ray inspection system 1 according to the present embodiment, the image generation unit 54 generates the defective product image G2 such that the X-ray attenuation rate in the virtual defect F is lower than the X-ray attenuation rate in the article A. Defects such as resin have a lower X-ray attenuation rate than the article A. Therefore, the server 50 can generate a defective product image G2 in which a virtual defect F such as resin is included inside the article A.

[0055] In the X-ray inspection system 1 according to this embodiment, the image generation unit 54 changes a plurality of adjacent pixel values in at least a part of the region corresponding to the article A. With this configuration, the pixel values in a continuous region are changed. Therefore, it is possible to appropriately generate a defective product image G2 including a virtual defect F such as resin inside the article A.

[0056] As described above, the embodiments of the present invention have been described. However, the present invention is not necessarily limited to the above-described embodiments, and various modifications are possible without departing from the gist thereof.

[0057] In the above embodiment, an example has been described in which the X-ray detection unit 16 of the X-ray inspection apparatus 10 includes one line sensor. However, the X-ray detection unit 16 may be capable of detecting X-rays in a specific energy band, or may be capable of detecting X-rays by a photon counting method. The X-ray detection unit 16 may be a direct conversion type detection unit or an indirect conversion type detection unit. The X-ray detection unit 16 may include, for example, a sensor (multi-energy sensor) that detects X-rays in each of a plurality of energy bands that penetrate the article A. The sensors are arranged, for example, in a direction (width direction) orthogonal to at least the conveyance direction and the vertical direction of the conveyance unit 14. The elements may be arranged not only in the width direction but also in the conveyance direction. That is, the X-ray detection unit 16 may include a line sensor or may include a sensor group arranged two-dimensionally. The sensor is, for example, a photon detection type sensor such as a CdTe semiconductor detector.

[0058] In the above embodiment, an example has been described in which the server 50 transmits a learned model to the X-ray inspection apparatus 10 in response to a request from the X-ray inspection apparatus 10. However, in the X-ray inspection apparatus 10, the learned model can also be acquired from an external storage medium such as a USB memory or a removable hard disk.

[0059] In the above embodiment, an example has been described in which the X-ray inspection system 1 includes the X-ray inspection apparatus 10 and the server 50. However, the X-ray inspection apparatus may include an image generation apparatus and a learned model generation apparatus.

[0060] In the above embodiment, an example has been described in which the server 50 has the functions of an image generation device and a learned model generation device. However, each of the image generation device and the learned model generation device may be configured in different servers or the like.

[0061] In the above embodiment, the X-ray inspection apparatus 10 has the control unit 18, but the present invention is not limited to this. For example, at least some functions of the control unit 18 may be implemented in an external control device (such as a laptop PC, a tablet, a server, etc.) capable of wired communication or wired communication with respect to the X-ray inspection apparatus 10. Each configuration in the above embodiment or the above modification example can be arbitrarily applied to each configuration in other embodiments or other modification examples.

Explanation of reference numerals

[0062] 10... X-ray inspection apparatus, 14... conveyance unit, 15... X-ray irradiation unit, 16... X-ray detection unit, 18... control unit (inspection unit), 53... image acquisition unit, 54... image generation unit, 55... learning unit, A... article, F... defect, G1... good product image (first image), G2... defective product image (second image).

Claims

1. An image acquisition unit that acquires a first image based on X-rays that have passed through an article without defects; An image generation unit that changes pixel values of at least a part of a region corresponding to the article in the first image acquired by the image acquisition unit, and generates a second image in which virtual defects are included in the article, the image generation apparatus comprising: The image generation unit generates the second image in which the virtual defects are included inside the article and the attenuation rate of the X-rays in the virtual defects is different from the attenuation rate of the X-rays in the article.

2. The image generation apparatus according to claim 1, wherein the image generation unit generates the second image such that the attenuation rate of the X-rays in the virtual defects is lower than the attenuation rate of the X-rays in the article.

3. The image generation apparatus according to claim 1 or 2, wherein the image generation unit changes a plurality of adjacent pixel values in at least a part of the region corresponding to the article.

4. An image generation apparatus according to claim 1 or 2; and A learning unit that performs machine learning using the second image generated by the image generation unit in the image generation apparatus as teacher data, and generates a learned model for determining whether or not the article includes the defect, the learned model generation apparatus comprising:

5. A learning unit that performs machine learning using the second image generated by the image generation unit in the image generation apparatus according to claim 1 or 2 as teacher data, and generates a learned model for determining whether or not the article includes the defect, the learned model generation apparatus comprising:

6. A conveyance unit that conveys an article; An X-ray irradiation unit that irradiates the article conveyed by the conveyance unit with X-rays; An X-ray detection unit that detects X-rays that have passed through the article after being irradiated by the X-ray irradiation unit; An inspection unit that generates an X-ray inspection image based on a detection result of the X-ray detection unit, and inspects the article based on the X-ray inspection image, the X-ray inspection apparatus comprising: The inspection unit determines whether or not the article includes the defect using the learned model generated by the learned model generation apparatus according to claim 4.

7. A conveyance unit that conveys an article; An X-ray irradiation unit that irradiates the article conveyed by the conveyance unit with X-rays; An X-ray detection unit that detects X-rays that have passed through the article after being irradiated by the X-ray irradiation unit; An inspection unit that generates an X-ray inspection image based on the detection result of the X-ray detection unit and inspects the article based on the X-ray inspection image. The inspection unit is an X-ray inspection apparatus that determines whether or not the article includes the defect using the learned model generated by the learned model generation apparatus according to claim 5.

8. The image generation apparatus according to claim 1 or 2, A learning unit that performs machine learning using the second image generated by the image generation unit in the image generation apparatus as teacher data and generates a learned model for determining whether or not the article includes the defect. A conveyance unit that conveys the article. An X-ray irradiation unit that irradiates the article conveyed by the conveyance unit with X-rays. An X-ray detection unit that detects the X-rays that have passed through the article after being irradiated by the X-ray irradiation unit. An inspection unit that generates an X-ray inspection image based on the detection result of the X-ray detection unit and inspects the article based on the X-ray inspection image. The inspection unit is an X-ray inspection apparatus that determines whether or not the article includes the defect using the learned model generated by the learning unit.

Images