Inspection processing device, and inspection processing method

Abstract

To provide an inspection processing device capable of preventing both overlooking and over-detection of abnormalities.SOLUTION: An inspection processing device 100 includes: a first learning model 102 that determines whether there are any abnormalities in an object of inspection that appear in an input image; a second learning model 103 that determines whether there are any abnormalities in an object of inspection that appear in an input image, which is a second learning model 103, with the performance of determining whether the test subject has an abnormality is lower than that of the first learning model 102; and a determination unit 104 that determines that there is an abnormality in the test subject when the first learning model 102 determines that there is an abnormality in the test subject and the second learning model 103 determines that the test subject is abnormal.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present disclosure relates to an inspection processing apparatus and an inspection processing method. [Background technology]

[0002] Patent Document 1 discloses a technique for inspecting foreign matter on a substrate. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2018-113467 Summary of the Invention [Problem to be solved by the invention]

[0004] In recent years, learning models have been used to inspect the presence or absence of abnormalities in inspection objects such as circuit boards. For example, when determining the presence or absence of abnormalities using a single learning model, it is necessary to set a strict anomaly judgment threshold (specifically, a low judgment threshold) to avoid overlooking abnormalities. However, in this case, non-abnormal areas may also be judged as abnormal. In other words, over-detection occurs. On the other hand, setting the judgment threshold too high may result in overlooking abnormalities.

[0005] Therefore, the present disclosure provides an inspection processing apparatus and the like that can suppress both overlooking abnormalities and over-detection. [Means for solving the problem]

[0006] The inspection processing device according to the present disclosure comprises a first learning model that determines whether or not an inspection object shown in an input image has an abnormality; a second learning model that determines whether or not an inspection object shown in the input image has an abnormality, the second learning model having lower performance than the first learning model in determining whether or not an abnormality exists in the inspection object; and a judgment unit that judges whether or not an abnormality exists in the inspection object when the first learning model determines that an abnormality exists in the inspection object and the second learning model determines that an abnormality exists in the inspection object.

[0007] These comprehensive or specific aspects may be realized by a system, an apparatus, a method, a recording medium, or a computer program, or may be realized by any combination of a system, an apparatus, a method, a recording medium, and a computer program. [Effects of the Invention]

[0008] According to the inspection processing device and the like according to the present disclosure, it is possible to prevent both overlooking and over-detection of abnormalities. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a configuration diagram illustrating an example of a component mounting system according to an embodiment. [Figure 2] 1 is a configuration diagram illustrating an example of a component mounting apparatus according to an embodiment; [Figure 3] 1 is a functional configuration diagram illustrating an example of an inspection processing apparatus according to an embodiment; [Figure 4] FIG. 10 is a diagram for explaining a learning method for each learning model. [Figure 5] 10 is a flowchart illustrating an example of an operation of the inspection processing apparatus according to the embodiment. [Figure 6] FIG. 10 is a diagram showing a two-dimensional map of the detection values ​​of each learning model. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, the embodiments will be specifically described with reference to the drawings.

[0011] The embodiments described below are all comprehensive or specific examples, and the numerical values, shapes, materials, components, arrangement and connection of the components, steps, and order of steps shown in the following embodiments are merely examples and are not intended to limit the present disclosure.

[0012] (Embodiment) An inspection processing apparatus according to an embodiment will be described below with reference to FIGS.

[0013] FIG. 1 is a configuration diagram showing an example of a component mounting system 1 according to an embodiment.

[0014] Component mounting system 1 includes a component mounting device that picks up components and mounts them on a board, and an inspection server 3 that inspects the inspection target. For example, while FIG. 1 shows an example in which component mounting system 1 includes two component mounting devices M1 and M2, component mounting system 1 may include only one component mounting device, or three or more. The following description focuses on component mounting device M1 of component mounting devices M1 and M2. Component mounting device M2 basically has the same configuration and functions as component mounting device M1, so a description of component mounting device M2 will be omitted.

[0015] For example, the inspection target is a circuit board during or before component mounting. For example, inspection server 3 inspects whether other components have fallen onto the circuit board at the component mounting position during component mounting. The inspection target may also be a tape feeder used for component mounting. For example, inspection server 3 inspects whether there are any components at the component removal position of the tape feeder before component mounting, whether there are any incorrect components at the component removal position, or whether the orientation of the components at the component removal position is misaligned. An example in which the inspection target is a circuit board will be described below. Component mounting devices M1 and M2 are connected to inspection server 3 via a communication network 2 such as a LAN (Local Area Network).

[0016] Next, the configuration of the component mounting apparatus M1 will be described with reference to FIG.

[0017] FIG. 1 is a configuration diagram showing an example of a component mounting apparatus M1 according to an embodiment.

[0018] In component mounting device M1, a board transport mechanism 5 is installed in the X direction at the center of base 4. Board transport mechanism 5 transports board B carried in from the upstream side in the X direction, and positions and holds it at a mounting operation position by mounting head 10, which will be described below. In addition, board transport mechanism 5 carries board B downstream after component mounting operation is completed. In other words, board B carried out from component mounting device M1 is carried to component mounting device M2.

[0019] A component supply unit 6 is installed on each side (front and back) of the board transport mechanism 5. A plurality of tape feeders 7 are attached in parallel in the X direction to the component supply units 6 on both sides. The tape feeders 7 feed a carrier tape, on which pockets for storing components are formed, by pitch feeding in a direction (tape feed direction) from the outside of the component supply unit 6 toward the board transport mechanism 5, thereby supplying components to a component removal position where the mounting head 10 picks up the components.

[0020] A Y-axis table 8 equipped with a linear drive mechanism is disposed on both ends of the upper surface of the base 4 in the X direction. A beam 9 similarly equipped with a linear drive mechanism is coupled to the Y-axis table 8 so as to be freely movable in the Y direction. A mounting head 10 is attached to the beam 9 so as to be freely movable in the X direction. The mounting head 10 is equipped with multiple suction units (not shown) that can hold and elevate components and rotate around a vertical axis (Z axis). A suction nozzle (not shown) that sucks and holds the component is attached to the bottom end of each suction unit. Note that, although both the Y-axis table 8 and the beam 9 are equipped with linear drive mechanisms here, this is not a limitation. For example, at least one of them may be a drive mechanism such as a ball screw.

[0021] The Y-axis table 8 and beam 9 constitute a mounting head moving mechanism 11 that moves the mounting head 10 horizontally (X direction, Y direction). The mounting head moving mechanism 11 and mounting head 10 perform a component mounting operation in which components are taken from the component supply unit 6 and mounted at the mounting positions on the board B. During the component mounting operation, the mounting head 10 moves above the component supply unit 6, picks up predetermined components with each suction nozzle, moves above the board B, rotates the components held by each suction nozzle in a predetermined direction, and repeats a series of turns to mount the components at their respective mounting positions.

[0022] A head camera 12 is attached to the beam 9, positioned on the underside of the beam 9 and moving integrally with the mounting head 10. As the mounting head 10 moves, the head camera 12 moves above the board B positioned at the mounting work position of the board transport mechanism 5, and captures an image of a board mark (not shown) provided on the board B to recognize the position of the board B. The head camera 12 also captures an image of the vicinity of the mounting position on the board B where components are to be mounted, and the state of the board B is recognized from the image capture results.

[0023] A component recognition camera 13 is installed between the component supply unit 6 and the board transport mechanism 5. When the mounting head 10, which has picked up a component from the component supply unit 6, is positioned above the component recognition camera 13, the component recognition camera 13 captures an image of the component held by the suction nozzle from below. During the component mounting operation by the mounting head 10, the mounting position is corrected taking into account the recognition results of the board B by the head camera 12 and the recognition results of the component by the component recognition camera 13.

[0024] A touch panel 14 operated by the worker is installed in front of component mounting device M1 at the position where the worker works. Touch panel 14 displays various information on its display, and the worker inputs data and operates component mounting device M1 using operation buttons and the like displayed on the display. A component disposal unit 15 is installed next to component recognition camera 13. Components that are supplied from component supply unit 6 but not mounted on board B, such as components that are poorly picked up, are discarded in component disposal unit 15.

[0025] For example, the inspection processing device according to the embodiment is provided in the inspection server 3. The inspection processing device may also be provided in the component mounting device M1. That is, the inspection processing device according to the embodiment may be provided in the component mounting device M1 where the inspection target exists, or may be provided in a device (e.g., the inspection server 3) outside the component mounting device M1 where the inspection target exists.

[0026] Next, the configuration of the inspection processing device will be described with reference to FIG.

[0027] FIG. 3 is a functional configuration diagram showing an example of the inspection processing apparatus 100 according to the embodiment.

[0028] The inspection processing device 100 is a device for inspecting whether an object to be inspected has an abnormality. For example, the object to be inspected is an object to be inspected in the component mounting device M1 (e.g., a substrate B or a tape feeder 7). The inspection processing device 100 includes a learning unit 101, a first learning model 102, a second learning model 103, and a determination unit 104. The learning unit 101, the first learning model 102, the second learning model 103, and the determination unit 104 are realized by a processor or the like that executes a program stored in a memory.

[0029] The learning unit 101 is a functional unit that causes learning to be performed on the first learning model 102 and the second learning model 103. Here, the learning method of each learning model will be described with reference to FIG.

[0030] FIG. 4 is a diagram for explaining the learning method of each learning model.

[0031] For example, in training each learning model, an abnormal image showing an abnormal inspection target is used as a correct image. Here, such an abnormal image is referred to as a defective image sample. For example, the first learning model 102 is a learning model trained with multiple defective image samples. Also, a reduced image obtained by reducing a defective image sample is used as a correct image. For example, a reduced image is an image in which the vertical size and horizontal size of a defective image sample are each reduced by half. For example, the second learning model 103 is a learning model trained with multiple reduced images obtained by reducing multiple defective image samples. In this way, by using the reduced images, the second learning model 103 with low performance in determining whether or not an abnormality exists can be trained. In other words, by using unreduced defective image samples, the first learning model 102 with high performance in determining whether or not an abnormality exists can be trained.

[0032] Also, for example, the first learning model 102 may be a learning model trained with a first neural network having multiple layers (e.g., 20 layers), and the second learning model 103 may be a learning model trained with a second neural network having fewer layers (e.g., 10 layers) than the first neural network. In this way, by using a second neural network with fewer layers, it is possible to train the second learning model 103 with low performance in determining whether or not an abnormality exists. In other words, by using a first neural network with many layers, it is possible to train the first learning model 102 with high performance in determining whether or not an abnormality exists.

[0033] Note that, if the second learning model 103 is a learning model trained with a second neural network having fewer layers than the first neural network, it does not necessarily have to be a learning model trained with a plurality of reduced images obtained by reducing a plurality of defective image samples. Also, if the second learning model 103 is a learning model trained with a plurality of reduced images obtained by reducing a plurality of defective image samples, it does not necessarily have to be a learning model trained with a second neural network having fewer layers than the first neural network. In other words, the second learning model 103 may be at least one of a learning model trained with a plurality of reduced images obtained by reducing a plurality of defective image samples and a learning model trained with a second neural network having fewer layers than the first neural network.

[0034] Furthermore, the inspection processing device 100 does not necessarily have to include the learning unit 101. For example, the first learning model 102 and the second learning model 103 learned by a learning device outside the inspection processing device 100 may be stored in the inspection processing device 100. Note that if the inspection processing device 100 includes the learning unit 101, each learning model can be updated using, for example, images acquired during inspection.

[0035] The determination unit 104 determines that the inspection object has an abnormality when the first learning model 102 determines that the inspection object has an abnormality and the second learning model 103 determines that the inspection object has an abnormality. The determination unit 104 also outputs a determination result indicating whether or not the inspection object has an abnormality. For example, the determination result may be output to a display or the like and displayed. The operation of the determination unit 104 will be described in detail with reference to FIG. 5.

[0036] FIG. 5 is a flowchart showing an example of the operation of the inspection processing apparatus 100 (specifically, the determining unit 104) according to the embodiment.

[0037] First, the determination unit 104 determines whether or not an abnormality exists in the inspection target using the first learning model 102 (step S11). Specifically, the determination unit 104 determines whether or not an abnormality exists in the inspection target by inputting an image of the inspection target into the first learning model 102.

[0038] The determination unit 104 determines whether a first detection value, which is an output of the first learning model 102 when an image is input to the first learning model 102, is equal to or greater than a first threshold (step S12). The first detection value is a feature of the inspection object obtained by the first learning model 102 and is normalized to a value between 0 and 1.0, for example. The first threshold is a threshold for determining whether or not an abnormality exists in the inspection object and is set appropriately. If the first detection value is equal to or greater than the first threshold, the determination using the first learning model 102 determines that an abnormality exists in the inspection object. In other words, if the first detection value is equal to or greater than the first threshold, the first learning model 102 can determine that an abnormality exists in the inspection object. Note that simply determining that an abnormality exists in the inspection object by the first learning model 102 does not ultimately determine that an abnormality exists in the inspection object, and processing in the next step S13 is performed.

[0039] If the first detection value is equal to or greater than the first threshold value (Yes in step S12), the determination unit 104 determines whether or not there is an abnormality in the inspection object using the second learning model 103 (step S13). Specifically, the determination unit 104 determines whether or not there is an abnormality in the inspection object by inputting an image showing the inspection object (the same image as the image input to the first learning model 102) into the second learning model 103.

[0040] The determination unit 104 determines whether a second detection value, which is an output of the second learning model 103 when an image is input to the second learning model 103, is equal to or greater than a second threshold (step S14). The second detection value is a feature of the inspection object obtained by the second learning model 103, and is normalized to a value between 0 and 1.0, for example. The second threshold is a threshold for determining whether or not an abnormality exists in the inspection object, and is set to a value smaller than the first threshold, for example. If the second detection value is equal to or greater than the second threshold, the determination using the second learning model 103 determines that an abnormality exists in the inspection object. In other words, if the second detection value is equal to or greater than the second threshold, the second learning model 103 can determine that an abnormality exists in the inspection object.

[0041] If the second detection value is equal to or greater than the second threshold value (Yes in step S14), the determination unit 104 determines that the inspection object has an abnormality (step S15). In this way, the determination unit 104 determines that the inspection object has an abnormality if the first detection value is equal to or greater than the first threshold value and the second detection value is equal to or greater than the second threshold value.

[0042] On the other hand, if the first detection value is less than the first threshold value (No in step S12) or if the second detection value is less than the second threshold value (No in step S14), the judgment unit 104 judges that the inspection object is a non-defective product (step S16).

[0043] Here, the detection values ​​of each learning model and the final determination result of whether or not an abnormality exists in the inspection object will be explained using FIG. 6.

[0044] Fig. 6 is a diagram showing a two-dimensional map of the detection values ​​of each learning model. The horizontal axis of the two-dimensional map shown in Fig. 6 indicates the first detection value, and the vertical axis indicates the second detection value. In Fig. 6, for each of the multiple inspection targets, the first detection value by the first learning model 102 and the second detection value by the second learning model 103 are represented by one point.

[0045] 6, it can be seen that an inspection object whose first detection value is equal to or greater than the first threshold and whose second detection value is equal to or greater than the second threshold is ultimately determined to have an abnormality (e.g., a defective product), whereas an inspection object whose first detection value is less than the first threshold or whose second detection value is less than the second threshold is ultimately determined to have no abnormality (e.g., a non-defective product).

[0046] As described above, since the presence or absence of an abnormality is determined by the second learning model 103, which has low performance in determining the presence or absence of an abnormality, it is possible to prevent the abnormality from being overlooked. Furthermore, since the presence or absence of an abnormality is determined by the first learning model 102, which has high performance in determining the presence or absence of an abnormality, it is possible to prevent over-detection of an abnormality. Therefore, by determining the presence or absence of an abnormality in the inspection object using both the first learning model 102 and the second learning model 103, it is possible to prevent both over-detection of an abnormality and over-detection.

[0047] (Other embodiments) The inspection processing apparatus 100 of the present disclosure has been described above based on the embodiment, but the present disclosure is not limited to the above embodiment. As long as it does not deviate from the spirit of the present disclosure, various modifications conceivable by a person skilled in the art to the present embodiment and forms constructed by combining components of different embodiments are also included within the scope of the present disclosure.

[0048] For example, in the above embodiment, the second threshold is smaller than the first threshold, but the second threshold may be equal to or larger than the first threshold.

[0049] For example, the present disclosure can be realized not only as the inspection processing device 100 but also as an inspection processing method including steps (processing) performed by the components that make up the inspection processing device 100.

[0050] In the inspection processing method, as shown in Figure 5, an image of the inspection object is input into a first learning model 102 to determine whether or not there is an abnormality in the inspection object (step S11), and the image is input into a second learning model 103, which has lower performance in determining whether or not there is an abnormality in the inspection object than the first learning model 102, to determine whether or not there is an abnormality in the inspection object (step S13).If it is determined that there is an abnormality in the inspection object based on the determination using the first learning model 102 (Yes in step S12) and if it is determined that there is an abnormality in the inspection object based on the determination using the second learning model 103 (Yes in step S14), it is determined that there is an abnormality in the inspection object (step S15).

[0051] For example, the present disclosure can be realized as a program for causing a computer (processor) to execute steps included in an inspection processing method. Furthermore, the present disclosure can be realized as a non-transitory computer-readable recording medium, such as a CD-ROM, on which the program is recorded.

[0052] For example, when the present disclosure is realized as a program (software), each step is performed by running the program using hardware resources such as a computer's CPU, memory, input / output circuits, etc. In other words, each step is performed by the CPU acquiring data from memory or input / output circuits, etc., performing calculations, and outputting the calculation results to memory or input / output circuits, etc.

[0053] In the above embodiment, each component included in the inspection processing device 100 may be configured with dedicated hardware, or may be realized by executing a software program suitable for each component. Each component may also be realized by a program execution unit such as a CPU or processor reading and executing a software program recorded on a recording medium such as a hard disk or semiconductor memory.

[0054] Some or all of the functions of the inspection processing device 100 according to the above embodiment are typically realized as an LSI, which is an integrated circuit. These may be individually integrated into single chips, or some or all of them may be integrated into a single chip. Furthermore, the integrated circuit is not limited to an LSI, and may be realized by a dedicated circuit or a general-purpose processor. An FPGA (Field Programmable Gate Array), which can be programmed after LSI manufacture, or a reconfigurable processor, which can reconfigure the connections and settings of circuit cells within an LSI, may also be used.

[0055] Furthermore, if an integrated circuit technology that can replace LSI emerges due to advances in semiconductor technology or other derivative technologies, it is natural that each component included in the inspection processing device 100 may be integrated using that technology.

[0056] In addition, this disclosure also includes forms obtained by making various modifications to the embodiments that a person skilled in the art would think of, and forms realized by arbitrarily combining the components and functions in each embodiment within the scope that does not deviate from the intent of this disclosure.

[0057] (Addendum) The above description of the embodiments discloses the following techniques.

[0058] (Technology 1) An inspection processing device comprising: a first learning model that determines whether or not an inspection object shown in an input image has an abnormality; a second learning model that determines whether or not an inspection object shown in the input image has an abnormality, the second learning model having lower performance than the first learning model in determining whether or not an abnormality exists in the inspection object; and a judgment unit that judges whether or not an abnormality exists in the inspection object when the first learning model judges that an abnormality exists in the inspection object and the second learning model judges that an abnormality exists in the inspection object.

[0059] According to this, since the presence or absence of an abnormality is determined by the second learning model, which has low performance in determining the presence or absence of an abnormality, it is possible to prevent the abnormality from being overlooked. Also, since the presence or absence of an abnormality is determined by the first learning model, which has high performance in determining the presence or absence of an abnormality, it is possible to prevent over-detection of an abnormality. Therefore, by determining the presence or absence of an abnormality of the inspection object using both such first learning model and second learning model, it is possible to prevent both over-detection of an abnormality and over-detection.

[0060] (Technology 2) The inspection processing device described in Technology 1, wherein the judgment unit judges that there is an abnormality in the inspection object when a first detection value, which is the output of the first learning model when the image is input to the first learning model, is equal to or greater than a first threshold, and a second detection value, which is the output of the second learning model when the image is input to the second learning model, is equal to or greater than a second threshold that is smaller than the first threshold.

[0061] According to this, when the first detection value is equal to or greater than the first threshold value, the first learning model can determine that there is an abnormality in the inspection object, and when the second detection value is equal to or greater than the second threshold value, the second learning model can determine that there is an abnormality in the inspection object.

[0062] (Technology 3) The inspection processing device described in Technology 1 or 2, wherein the first learning model is a learning model trained on a plurality of abnormal images, and the second learning model is a learning model trained on a plurality of reduced images obtained by reducing the plurality of abnormal images.

[0063] According to this, by using reduced images, it is possible to train a second learning model with low performance in determining whether or not there is an abnormality. In other words, by using non-reduced abnormal images, it is possible to train a first learning model with high performance in determining whether or not there is an abnormality.

[0064] (Technology 4) An inspection processing device described in any one of Technologies 1 to 3, wherein the first learning model is a learning model trained with a first neural network having multiple layers, and the second learning model is a learning model trained with a second neural network having fewer layers than the first neural network.

[0065] According to this, by using a second neural network with fewer layers, it is possible to train a second learning model with lower performance in determining whether or not there is an abnormality. In other words, by using a first neural network with more layers, it is possible to train a first learning model with higher performance in determining whether or not there is an abnormality.

[0066] (Technology 5) The inspection processing device according to any one of technologies 1 to 4, further comprising a learning unit that causes the first learning model and the second learning model to learn.

[0067] In this way, the inspection processing device may be equipped with a learning unit, and each learning model can be updated using, for example, images acquired during inspection.

[0068] (Technical Aspect 6) The inspection processing apparatus according to any one of Technical Aspects 1 to 5, wherein the inspection object is an inspection object in a component mounting apparatus that picks up components and mounts them on a board.

[0069] For example, the object to be inspected may be a substrate or a tape feeder in a component mounting device.

[0070] (Technology 7) An inspection processing method in which an image of an inspection object is input into a first learning model to determine whether or not there is an abnormality in the inspection object, and the image is input into a second learning model that has lower performance than the first learning model in determining whether or not there is an abnormality in the inspection object to determine whether or not there is an abnormality in the inspection object, and if the first learning model determines that there is an abnormality in the inspection object and the second learning model determines that there is an abnormality in the inspection object, the method determines that there is an abnormality in the inspection object.

[0071] This makes it possible to provide an inspection processing method that can prevent both overlooking and over-detection of abnormalities. [Industrial Applicability]

[0072] The present disclosure can be applied to a component mounting system in which an inspection target is inspected in a component mounting device. [Explanation of symbols]

[0073] 1. Component mounting system 2. Communication Network 3 Inspection Server 4 Foundation 5. Substrate transport mechanism 6. Parts Supply Department 7 Tape Feeder 8 Y-axis table 9 Beam 10 Mounting head 11 Mounting head movement mechanism 12 Head Camera 13 Parts Recognition Camera 14 Touch Panel 15 Parts Disposal Department 100 Inspection processing device 101 Learning Department 102 First Learning Model 103 Second Learning Model 104 Judgment section B board M1, M2 component mounting equipment

Claims

1. a first learning model that determines whether or not an abnormality exists in an inspection object captured in an input image; a second learning model that determines whether or not an abnormality exists in the inspection object captured in the input image, the second learning model having a lower performance in determining whether or not an abnormality exists in the inspection object than the first learning model; a determination unit that determines that the inspection object has an abnormality when the first learning model determines that the inspection object has an abnormality and the second learning model determines that the inspection object has an abnormality, Inspection processing equipment.

2. the determination unit determines that the inspection object has an abnormality when a first detection value, which is an output of the first learning model when the image is input to the first learning model, is equal to or greater than a first threshold, and when a second detection value, which is an output of the second learning model when the image is input to the second learning model, is equal to or greater than a second threshold that is smaller than the first threshold; The inspection processing apparatus according to claim 1 .

3. the first learning model is a learning model trained using a plurality of abnormal images, The second learning model is a learning model trained using a plurality of reduced images obtained by reducing the plurality of abnormal images. The inspection processing apparatus according to claim 1 .

4. the first learning model is a learning model trained by a first neural network having a plurality of layers; the second learning model is a learning model trained by a second neural network having fewer layers than the first neural network; The inspection processing apparatus according to claim 1 .

5. Further, a learning unit that causes the first learning model and the second learning model to learn. The inspection processing apparatus according to claim 1 .

6. The inspection object is an inspection object in a component mounting device that picks up components and mounts them on a board. The inspection processing apparatus according to any one of claims 1 to 5.

7. inputting an image of an inspection object into a first learning model to determine whether or not an abnormality exists in the inspection object; determining whether or not the inspection object has an abnormality by inputting the image into a second learning model having a lower performance in determining whether or not the inspection object has an abnormality than the first learning model; determining that the inspection object has an abnormality when the determination using the first learning model is used and when the determination using the second learning model is used; and Inspection processing method.

Images