Detection model generation device, inspection device, and detection model generation method

Abstract

This invention provides a method for detecting objects that have not been previously inspected using X-ray inspection. [Solution] The detection model generation device includes means for constructing an X-ray image generation model that learns the relationship between an X-ray image taken during inspection by an X-ray inspection device and information about an object contained in the X-ray image, and outputs an X-ray image of the object when information about the object is input; means for receiving information about a new object that has not been inspected by the X-ray inspection device; means for inputting the information about the new object into the X-ray image generation model and obtaining an X-ray image of the new object output by the X-ray image generation model; and means for constructing an object detection model that learns the relationship between an X-ray image and an object contained in the X-ray image, and outputs the type of the object when the X-ray image is input. The means for constructing the object detection model constructs the object detection model so that it outputs the type of the new object when an X-ray image of the new object is input.

Description

Technical Field

[0001] The present disclosure relates to a detection model generation device, an inspection device, and a detection model generation method.

Background Art

[0002] In non-combustible waste treatment facilities, the introduction of a system for performing X-ray inspections to detect combustible materials (inappropriate materials) such as batteries mixed in non-combustible waste is being considered. Among non-combustible waste, smartphones, electric shavers, etc. are updated annually in terms of models, and their designs and shapes are changing. In a system for detecting specific objects by X-ray inspection, it is often the case that the inspection target is detected using X-ray transmission images acquired during past inspections. Therefore, the objects that can be detected are limited to those that have been inspection targets in the past. Therefore, when a new product that has never been the target of an X-ray inspection is processed as non-combustible waste, there is a possibility that the product cannot be detected. As a result, there is a risk that a product containing a flammable object such as a battery will be mixed into a crusher, leading to a large-scale disaster. Patent Document 1 discloses a technique for improving the accuracy of image recognition in X-ray inspection operations.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] There is a need for a technique that enables the detection of objects that have never been inspection targets by X-ray inspection. Patent Document 1 does not disclose a technique for dealing with such problems.

[0005] The present disclosure provides a detection model generation device, an inspection device, and a detection model generation method that can solve the above problems.

Means for Solving the Problems

[0006] The detection model generation device of this disclosure includes means for constructing an X-ray image generation model that learns the relationship between an X-ray image taken during inspection using an X-ray inspection device and information about an object contained in the X-ray image, and outputs an X-ray image of the object when information about the object is input; means for receiving information about a new object that has not been inspected by the X-ray inspection device; means for inputting the information about the new object into the X-ray image generation model and obtaining an X-ray image of the new object output by the X-ray image generation model; and means for constructing a detection model that learns the relationship between an X-ray image and an object contained in the X-ray image, and outputs the type of the object when the X-ray image is input, wherein the means for constructing the detection model constructs the detection model so that it outputs the type of the new object when an X-ray image of the new object is input.

[0007] The inspection apparatus of this disclosure includes means for detecting objects included in an X-ray image by inputting an X-ray image taken during inspection by the X-ray inspection apparatus into the object detection model generated by the detection model generation apparatus, and means for issuing an alarm if the detected object is a predetermined unsuitable object.

[0008] The detection model generation method of this disclosure includes the steps of: constructing an X-ray image generation model that learns the relationship between an X-ray image taken during inspection using an X-ray inspection device and information about an object contained in the X-ray image, and outputs an X-ray image of the object when information about the object is input; receiving information about a new object that has not been inspected by the X-ray inspection device; inputting the information about the new object into the X-ray image generation model and obtaining an X-ray image of the new object output by the X-ray image generation model; and constructing a detection model that learns the relationship between an X-ray image and an object contained in the X-ray image, and outputs the type of the object when the X-ray image is input, wherein in the step of constructing the detection model, the detection model is constructed to output the type of the new object when the X-ray image of the new object is input. [Effects of the Invention]

[0009] According to the detection model generation apparatus, inspection apparatus, and detection model generation method of this disclosure, it is possible to detect objects that have never been inspected before by X-ray inspection. [Brief explanation of the drawing]

[0010] [Figure 1] This is a schematic diagram of the inspection system according to the first embodiment. [Figure 2] This diagram shows the data flow in the inspection system according to the first embodiment. [Figure 3] This figure shows an example of the operation of the inspection system according to the first embodiment. [Figure 4] This figure shows an example of an X-ray image and context according to the first embodiment. [Figure 5] This is a schematic diagram of the inspection system according to the second embodiment. [Figure 6] This diagram shows the data flow in the inspection system according to the second embodiment. [Figure 7] This figure shows an example of the operation of the inspection system according to the second embodiment. [Figure 8] This figure shows an example of the hardware configuration of the control device according to the embodiment. [Modes for carrying out the invention]

[0011] <First Embodiment> The inspection system described in this disclosure will be explained below with reference to the drawings. (composition) Figure 1 is a schematic diagram of the inspection system according to the first embodiment. Figure 1 is a top view of the inspection system 100. The inspection system 100 is introduced, for example, into a non-combustible waste treatment facility to detect unsuitable materials mixed in with the waste. As shown in Figure 1, the inspection system 100 comprises an X-ray inspection device 1, a support device 20, and a belt conveyor 4. The X-ray inspection device 1 inspects the contents of transparent or translucent garbage bags 5 that are transported by the belt conveyor 4. The X-ray inspection device 1 comprises an X-ray camera 2, a visible light camera 3, and an inspection device 10. The X-ray camera 2 and the visible light camera 3 are connected to the inspection device 10. The X-ray camera 2 includes an X-ray emitter and an X-ray detector (such as a line sensor), and irradiates the garbage bags 5 passing through the X-ray inspection device 1 with X-rays, capturing an X-ray transmission image (referred to as an X-ray image). The X-ray image is transmitted to the inspection device 10. The visible light camera 3 captures a visible image of the garbage bags 5 passing through the X-ray inspection device 1. The visible image is transmitted to the inspection device 10. The X-ray camera 2 and the visible camera 3 are pointed in the same direction and positioned to photograph the garbage bag 5 from the same direction.

[0012] The inspection device 10 comprises an X-ray image acquisition unit 11, a visible image acquisition unit 12, and a detection unit 13. The X-ray image acquisition unit 11 acquires X-ray images as they pass through the X-ray inspection device 1. The X-ray images may be either directly captured black and white X-ray images or color images color-coded according to material. The visible image acquisition unit 12 acquires a visible image of an object passing through the X-ray inspection device 1.

[0013] The detection unit 13 determines whether or not the garbage bag 5 passing through the X-ray inspection device 1 contains unsuitable items such as lithium-ion batteries. For example, if a smartphone or electric shaver containing a lithium-ion battery is found in the garbage bag 5, the detection unit 13 issues an alarm. The detection unit 13 has a detection model 131 that processes X-ray images. The detection model 131 is a deep learning model such as a CNN or Transformer. The detection unit 13 has the function of constructing the detection model 131 using machine learning or deep learning. When the detection unit 13 receives an X-ray image as input, it learns the detection model 131 to output the position (coordinate information of the object in the X-ray image) and type of object contained in the garbage bag 5. The detection unit 13 has an unsuitable item list in which the types of unsuitable items are registered. If the type of object output by the detection model 131 is registered in the unsuitable item list, the detection unit 13 determines that an unsuitable item is present and issues an alarm.

[0014] The support device 20 generates new object information by combining an X-ray image of an object that has never been inspected before (referred to as a new object) with information about that object (type, etc.), and provides it to the inspection device 10. The inspection device 10 and the support device 20 are connected in a communication manner. In the inspection device 10, the detection unit 13 learns the new object information as training data and updates the detection model 131 so that it can detect new objects. The support device 20 includes a context generation unit 21, an image generation AI 22, and a storage unit 23.

[0015] The context generation unit 21 generates a context corresponding to the visible image captured by the X-ray inspection apparatus 1. The context includes the type of the object (such as a smartphone, a battery, etc.) and the position information (coordinate information) of the object included in the visible image. The context may be generated by software that outputs the context when a visible image is input, for example, a text generation AI such as an LLM (Large Language Model), etc., or a person may confirm the visible image and generate the context. When generated by software, the context generation unit 21 may include such software, or may acquire the context generated by the software. When a person generates the context, the context generation unit 21 acquires the generated context.

[0016] The image generation AI 22 acquires the visible image and context of the new object, and inputs this information into the image generation model 223 (FIG. 2) to acquire the X-ray image of the new object. The image generation model 223 is a machine learning model constructed by learning the relationship between the visible image, the X-ray image, and the context. The image generation AI 22 has a function of constructing the image generation model 223 by machine learning, deep learning, etc. The image generation AI 22 learns the image generation model 223 so that when the visible image and context of an object are input, the X-ray image of that object is output. When the image generation AI 22 generates the X-ray image of the new object using the image generation model, it outputs the X-ray image of the new object and the context of the new object to the inspection apparatus 10 (detection unit 13). The detection unit 13 learns the X-ray image of the new object and the context of the new object, and updates the detection model 131, so that the new object can be detected from the X-ray image of the garbage bag 5 passing through the X-ray inspection apparatus 1.

[0017] The storage unit 23 stores the X-ray image captured by the X-ray camera 2, the visible image captured by the visible camera 3, the context generated by the context generation unit 21, the X-ray image generated by the image generation AI 22, etc.

[0018] FIG. 2 shows the data flow in the inspection system 100 and the more detailed configurations of the inspection apparatus 10 and the support apparatus 20. The X-ray image acquisition unit 11 acquires an X-ray image and outputs the X-ray image to the support device 20. The support device 20 acquires the X-ray image and stores the X-ray image in the X-ray image storage unit 232 included in the storage unit 23. The visible image acquisition unit 12 acquires a visible image and outputs the visible image to the support device 20. The support device 20 acquires the visible image and stores the visible image in the visible image storage unit 231 included in the storage unit 23. The context generation unit 21 acquires the visible image stored in the visible image storage unit 231 and generates a context (the type of object included in the visible image and the coordinate information of the object). The context generation unit 21 stores the generated context in the context storage unit 233 included in the storage unit 23. The X-ray image, visible image, and context of the same object may be stored in the respective storage units 231 to 233 in association with identification information generated by any logic so that their correspondence relationships can be understood.

[0019] The image generation AI 22 includes a source context input unit 221, a source visible image input unit 222, and an image generation model 223. The image generation AI 22 acquires the visible image, X-ray image, and context related to one object from the visible image storage unit 231, X-ray image storage unit 232, and context storage unit 233, respectively, and uses these as one dataset. The image generation AI 22 creates a plurality of such datasets and trains the image generation model 223 to output an X-ray image when a visible image and a context are input. The datasets used for training are not data related to new objects but data of objects with inspection records.

[0020] In response, the user obtains a visible image of a new object that has not been handled by the inspection system 100, generates a context, and inputs the visible image and context of the new object to the support device 20. The source context input unit 221 accepts the input of the context of the new object (the context corresponding to the newly generated X-ray image). The source visible image input unit 222 accepts the input of the visible image of the new object (the visible image corresponding to the newly generated X-ray image). The visible image may be an image taken with the user's camera, or an image obtained from the internet, etc. The context generation unit 21 may also be used to generate the context of the new object.

[0021] The source context input unit 221 inputs the context of the new object to the image generation model 223. The source visible image input unit 222 inputs the visible image of the new object to the image generation model 223. The image generation model 223 generates an X-ray image of the new object from this information. The image generation AI 22 stores the generated X-ray image of the new object in the generated X-ray image storage unit 234 of the storage unit 23. The source context input unit 221 also stores the context of the new object input to the image generation model 223 in the generated X-ray context storage unit 235 of the storage unit 23.

[0022] The detection unit 13 acquires the X-ray image and context of a new object from the generated X-ray image storage unit 234 and the generated X-ray context storage unit 235, respectively. When this X-ray image is input, the detection model 131 is trained to output the context of the new object. As a result, when the garbage bag 5 containing the new object passes through the X-ray inspection device 1, the detection unit 13 can detect the new object from the X-ray image taken at that time.

[0023] Figure 3 shows an example of an X-ray image and context. An example of an X-ray image is shown on the left side of Figure 3. The X-ray image 30 includes a smartphone 31 and a battery 32. On the right side of Figure 3, the context 34 corresponding to the X-ray image 30 is shown. Context 34 includes the type and location information of new objects contained in the X-ray image 30. The four corners of the smartphone 31 and battery 32 are denoted as P1, P2, P3, and P4. P1 to P4 in context 34 represent the coordinate information of the four corners of the smartphone 31 and battery 32. For example, when the X-ray image 30 is input to the detection model 131, the detection model 131 outputs the context 34.

[0024] Figure 4 shows the processing flow for each configuration of the inspection system 100. The processing shown in Figure 4 is for constructing a detection model 131 that detects new objects, and the processing is executed in the order of B1 to B4 described below.

[0025] Figure 4B1 shows the on-site image acquisition process. The on-site image acquisition process is the process of saving information (X-ray image, visible image, context) of an object that has passed through the X-ray inspection device 1. The X-ray inspection device 1 acquires X-ray images and visible images, and the inspection device 10 stores these images in memory or the like. The X-ray inspection device 1 outputs the X-ray images and visible images to the support device 20, and the storage unit 23 stores them. In the support device 20, the context generation unit 21 acquires the visible image from the storage unit 23 and generates a context. The context generation unit 21 saves the generated context to the storage unit 23.

[0026] Figure 4B2 shows the update process for the generating AI. The update process for the generating AI is the update process for the image generation model 223. The image generation AI 22 obtains the corresponding X-ray image, visible image, and context from the memory unit 23, learns their relationships, and updates the image generation model 223 by further learning so that it can generate an X-ray image from the visible image and context.

[0027] Figure 4B3 shows the X-ray image generation process using the AI. The X-ray image generation process generates an X-ray image of a new object. The user inputs a visible image and context of the new object to the support device 20. The image generation AI 22 generates an X-ray image of the new object from the input visible image and context. The image generation AI 22 stores the X-ray image and context of the new object in the storage unit 23.

[0028] Figure 4B4 shows the process of updating the detection model. The X-ray inspection device 1 (detection unit 13) acquires the X-ray image and context of a new object from the memory unit 23, learns the relationship between them, and updates the detection model 131 by further learning so that it can generate the context of a new object when the X-ray image of a new object is input.

[0029] When the detection model 131 is updated, the detection unit 13 can detect if a new object is an unsuitable object based on the updated detection model 131 and the list of unsuitable objects.

[0030] (effect) As described above, according to this embodiment, if a visible image of a new object can be obtained, an X-ray image of the new object that has not previously passed through the X-ray inspection device 1 can be generated from the visible image of the new object and context information. Then, the relationship between the X-ray image of the new object and the context is learned. This makes it possible to detect new objects from X-ray images and prevent the inclusion of unsuitable substances in new products. In the above description, the inspection device 10 and the support device 20 are described as separate units, but they may be implemented in a single computer. Also, although the detection model 131 is configured to output object type and location information when an X-ray image is input, the detection model 131 may be configured to output only one of the object type or location information when an X-ray image is input.

[0031] <Second Embodiment> The X-ray inspection apparatus 1 of the first embodiment is equipped with a visible camera 3, and an image generation model 223 is trained to generate an X-ray image of a new object from the visible image and context of the new object using the visible image captured by the visible camera 3. However, there are X-ray inspection apparatuses that do not have a visible camera 3. In the second embodiment, a system is provided that can detect new objects in such X-ray inspection apparatuses.

[0032] (composition) Figure 5 is a schematic diagram of the inspection system according to the second embodiment. The inspection system 100a comprises an X-ray inspection device 1a, a support device 20a, and a belt conveyor 4. The X-ray inspection device 1a comprises an X-ray camera 2 and an inspection device 10a. The X-ray camera 2 is connected to the inspection device 10a. The X-ray camera 2 irradiates the garbage bags 5 passing through the X-ray inspection device 1a with X-rays and takes an X-ray image. The X-ray image is transmitted to the inspection device 10a.

[0033] The inspection device 10a comprises an X-ray image acquisition unit 11 and a detection unit 13. The inspection device 10a differs from the inspection device 10 of the first embodiment in that it does not have a visible image acquisition unit 12, but the other configurations are the same as those of the first embodiment. That is, the X-ray image acquisition unit 11 acquires an X-ray image (which may be a black and white X-ray image or a color image color-coded by material) passing through the X-ray inspection device 1a. The detection unit 13 uses the X-ray image acquired by the X-ray image acquisition unit 11 and the detection model 131 to determine whether or not the garbage bag 5 passing through the X-ray inspection device 1a contains unsuitable material.

[0034] The support device 20a comprises a context generation unit 21a, an image generation AI 22a, and a storage unit 23a. The context generation unit 21a generates a context for an object that has passed through the X-ray inspection device 1a. The context includes the type of object (smartphone, battery, etc.) and the object's position information (coordinate information) included in the X-ray image. For example, the context generation unit 21 may be generated by software that outputs a context when an X-ray image (or visible image) is input. Alternatively, the user may generate a context for an object that has passed through the X-ray inspection device 1a and input it to the context generation unit 21.

[0035] The image generation AI 22a acquires a context and inputs it into the image generation model 223a (Figure 6) to obtain an X-ray image of a new object. The image generation model 223a is a machine learning model built by learning the relationship between X-ray images and context. The image generation AI 22a has the function of building the image generation model 223a using machine learning, deep learning, etc. When the image generation AI 22a inputs the context of an object, it trains the image generation model 223a to output an X-ray image of the object. When the image generation AI 22a inputs the context of a new object into the image generation model 223a and generates an X-ray image of the new object, it outputs the X-ray image of the new object and the context of the new object to the inspection device 10a (detection unit 13). The detection unit 13 updates the detection model 131 by learning the X-ray image of the new object and the context of the new object. This makes it possible to detect new objects from the X-ray image of the garbage bag 5 passing through the X-ray inspection device 1a.

[0036] The memory unit 23a stores X-ray images taken by the X-ray camera 2, the context of new objects, and X-ray images generated by the image generation AI 22a.

[0037] Figure 6 shows the data flow in the inspection system 100a and a more detailed configuration of the inspection device 10a and the support device 20a. The X-ray image acquisition unit 11 acquires an X-ray image and outputs it to the support device 20a. The support device 20a acquires the X-ray image and stores it in the X-ray image storage unit 232 provided in the storage unit 23a. The context generation unit 21a generates context (type of object included in the X-ray image and coordinate information of the object) of an object that has passed through the X-ray inspection device 1a. The context generation unit 21a stores the generated context in the context storage unit 233 provided in the storage unit 23a. X-ray images and contexts of the same object may be stored in their respective memory units 232-233, associated with identification information generated by arbitrary logic, for example, so that their correspondence with each other can be understood.

[0038] The image generation AI 22a comprises a source context input unit 221 and an image generation model 223. The image generation AI 22a acquires an X-ray image and context for a single object from the X-ray image storage unit 232 and the context storage unit 233, respectively, and combines these into a single dataset. The image generation AI 22a creates multiple such datasets and trains the image generation model 223a to output an X-ray image when context is input. The datasets used for training are not data related to new objects, but data of objects that have been inspected before.

[0039] In response, the user generates a context for a new object that has not been handled by the inspection system 100a and inputs the context of the new object to the support device 20a. For example, the user may obtain an X-ray image or visible image of a new object taken at another non-combustible waste treatment facility and generate a context for the new object shown in the image. The source context input unit 221 accepts input of the context of the new object. Alternatively, the context generation unit 21a may be used to generate the context of the new object. For example, the obtained X-ray image or visible image of the new object may be input to the context generation unit 21a, and the context generation unit 21a may generate a context from the visible image of the new object.

[0040] The source context input unit 221 inputs the context of the new object to the image generation model 223a. The image generation model 223a generates an X-ray image of the new object. The image generation AI 22a saves the generated X-ray image of the new object to the generated X-ray image storage unit 234 provided in the storage unit 23a. The source context input unit 221 also saves the context of the new object input to the image generation model 223 to the generated X-ray context storage unit 235 provided in the storage unit 23a.

[0041] The detection unit 13 acquires the X-ray image and context of a new object from the generated X-ray image storage unit 234 and the generated X-ray context storage unit 235, respectively. When this X-ray image is input, the detection model 131 learns to output the context of the new object. As a result, the detection unit 13 can detect a new object from the X-ray image taken when the new object passes through the X-ray inspection device 1.

[0042] Figure 7 shows the processing flow for each configuration of the inspection system 100a. The processing shown in Figure 7 is for constructing a detection model 131 that detects a new object according to the second embodiment, and the processing is executed in the order of B1a to B4a described below.

[0043] Figure 7B1a shows the on-site image acquisition process. The X-ray inspection device 1a acquires and stores X-ray images. The X-ray inspection device 1a outputs the X-ray images to the support device 20a, and the storage unit 23a stores them. In the support device 20a, the context generation unit 21a generates context for objects included in the X-ray images. The context generation unit 21a stores the generated context in the storage unit 23a.

[0044] Figure 7, B2a shows the update process for the generated AI. The image generation AI 22a obtains the corresponding X-ray image and context from the memory unit 23a, uses these as training data, and updates the image generation model 223a by further training so that it can generate an X-ray image from the context.

[0045] Figure 7B3a shows the X-ray image generation process using the AI. The user inputs the context of a new object into the support device 20a. The image generation AI 22a uses the image generation model 223a to generate an X-ray image of the new object from the input context. The image generation AI 22a saves the X-ray image of the new object and its context to the storage unit 23a.

[0046] Figure 4B4a shows the process of updating the detection model. The X-ray inspection device 1a (detection unit 13) acquires the X-ray image and context of a new object from the storage unit 23a, uses these as training data, and updates the detection model 131 by further training so that when the X-ray image of a new object is input, it can generate the context of the new object.

[0047] (effect) As described above, according to this embodiment, it is possible to generate an X-ray image of a new object that has not previously passed through the X-ray inspection device 1a, based on the context of the new object. Then, the relationship between the X-ray image of the new object and the context is learned. This makes it possible to detect new objects from the X-ray image and prevent the inclusion of unsuitable substances in new products.

[0048] Figure 8 shows an example of the hardware configuration of the control device. The computer 900 includes a CPU 901, main memory 902, auxiliary memory 903, input / output interface 904, and communication interface 905. The inspection devices 10, 10a and support devices 20, 20a described above are implemented in the computer 900. The functions described above are stored in the auxiliary memory 903 in the form of programs. The CPU 901 reads the program from the auxiliary memory 903, expands it into the main memory 902, and executes the above processing according to the program. The CPU 901 also allocates memory space in the main memory 902 according to the program. The CPU 901 also allocates memory space in the auxiliary memory 903 to store the data being processed according to the program.

[0049] Furthermore, a program to implement all or part of the functions of the inspection devices 10, 10a and the support devices 20, 20a may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be loaded into a computer system and executed to perform processing by each functional unit. Here, "computer system" includes hardware such as the OS and peripheral devices. Also, if a WWW system is used, "computer system" also includes the homepage provisioning environment (or display environment). Furthermore, "computer-readable recording medium" refers to portable media such as CDs, DVDs, USBs, and storage devices such as hard disks built into the computer system. Furthermore, if this program is distributed to computer 900 via a communication line, computer 900 that receives the distribution may load the program into main memory 902 and execute the above processing. Moreover, the above program may be for implementing only a part of the functions described above, and may also be able to implement the above functions in combination with programs already recorded in the computer system.

[0050] As described above, several embodiments relating to this disclosure have been explained, but all of these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be carried out in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims and their equivalents.

[0051] <Note> The detection model generation apparatus, inspection apparatus, and detection model generation method described in the embodiment can be understood, for example, as follows.

[0052] (1) The detection model generation device according to the first embodiment includes means for constructing an X-ray image generation model that learns the relationship between an X-ray image taken during inspection using an X-ray inspection device and information of an object contained in the X-ray image, and outputs an X-ray image of the object when information of the object is input; means for receiving information of a new object that has not been inspected by the X-ray inspection device; means for inputting the information of the new object into the X-ray image generation model and obtaining an X-ray image of the new object output by the X-ray image generation model; and means for constructing a detection model that learns the relationship between an X-ray image and an object contained in the X-ray image, and outputs the type of the object when the X-ray image is input, wherein the means for constructing the detection model constructs the detection model so as to output the type of the new object when an X-ray image of the new object is input. This makes it possible to detect objects that have never been inspected before using X-ray inspection.

[0053] (2) The detection model generation device according to the second embodiment is the detection model generation device of (1), wherein the object information is the type of the object and the position information of the object in the X-ray image, and the new object information is the type of the new object and the position information of the object in the X-ray image. This makes it possible to detect new objects using the method of the second embodiment.

[0054] (3) The detection model generation device according to the third embodiment is the detection model generation device according to (1) to (2), wherein the information of the object is a visible image of the object, the type of the object, and the position information of the object in the visible image, and the information of the new object is a visible image of the new object, the type of the new object, and the position information of the new object in the visible image. This makes it possible to detect new objects using the method of the first embodiment.

[0055] (4) A detection device according to the fourth embodiment, comprising means for detecting an object included in an X-ray image by inputting an X-ray image taken during inspection by the X-ray inspection device into the detection model generated by the detection model generation device of (1) to (3), and means for issuing an alarm when the detected object is a predetermined unsuitable object. This makes it possible to detect unsuitable objects passing through the X-ray inspection device.

[0056] (5) The detection model generation method according to the fifth embodiment includes the steps of: constructing an X-ray image generation model that learns the relationship between an X-ray image taken during inspection using an X-ray inspection device and information about an object contained in the X-ray image, and outputs an X-ray image of the object when information about the object is input; receiving information about a new object that has not been inspected with the X-ray inspection device; inputting the information about the new object into the X-ray image generation model and obtaining an X-ray image of the new object output by the X-ray image generation model; and constructing a detection model that learns the relationship between an X-ray image and an object contained in the X-ray image, and outputs the type of the object when the X-ray image is input, wherein in the step of constructing the detection model, the detection model is constructed to output the type of the new object when an X-ray image of the new object is input. [Explanation of symbols]

[0057] 1, 1a...X-ray inspection equipment 2. X-ray camera 3. Visible Camera 4. Belt conveyor 5. Garbage bags 10, 10a... Inspection device 11...X-ray image acquisition section 12. Visible Image Acquisition Unit 13. Detection Unit 131...Detection Model 20, 20a... Support device 21, 21a...Context generation unit 22, 22a...Image generation AI 23, 23a...Storage section 100, 100a... Inspection system 900... Computer 901···CPU 902...Main memory 903...Auxiliary storage device 904... Input / Output Interface 905...Communication Interface

Claims

1. A means for constructing an X-ray image generation model that learns the relationship between an X-ray image taken during inspection using an X-ray inspection device and the information of an object contained in the X-ray image, and outputs an X-ray image of the object when the information of the object is input, A means for receiving information on a new object that has not been inspected by the aforementioned X-ray inspection device, A means for inputting information about the new object into the X-ray image generation model and obtaining an X-ray image of the new object output by the X-ray image generation model, A means for constructing a detection model that learns the relationship between an X-ray image and the objects contained in the X-ray image, and outputs the type of object when the X-ray image is input, Equipped with, The means for constructing the detection model constructs the detection model so that when an X-ray image of the new object is input, it outputs the type of the new object. Detection model generation device.

2. The information about the object includes the type of the object and the position information of the object in the X-ray image. The information of the new object includes the type of the new object and the position information of the object in the X-ray image. The detection model generation device according to claim 1.

3. The information of the object includes a visible image of the object, the type of the object, and the position information of the object in the visible image. The information of the new object includes a visible image of the new object, the type of the new object, and the position information of the new object in the visible image. The detection model generation device according to claim 1.

4. A means for detecting an object included in an X-ray image by inputting an X-ray image taken during inspection by the X-ray inspection device into the detection model generated by the detection model generation device according to any one of claims 1 to 3, A means for issuing an alarm if the detected object is a predetermined unsuitable object, An inspection device equipped with the following features.

5. The steps include: constructing an X-ray image generation model that learns the relationship between an X-ray image taken during inspection using an X-ray inspection device and the information of an object contained in the X-ray image, and outputs an X-ray image of the object when the information of the object is input; The steps include receiving information on a new object that has not been inspected by the aforementioned X-ray inspection device, The steps include inputting information about the new object into the X-ray image generation model and obtaining an X-ray image of the new object output by the X-ray image generation model, The steps include: constructing a detection model that learns the relationship between an X-ray image and the objects contained in the X-ray image, and outputs the type of object when the X-ray image is input; It has, In the step of constructing the detection model, the detection model is constructed such that when an X-ray image of the new object is input, the type of the new object is output. Detection model generation method.

Images