Product inspection device

Abstract

To provide an item inspection device that enables efficient image processing without requiring an additional power supply. [Solution] The item inspection device 1 includes an image storage unit 4 that stores inspection images obtained by imaging the item to be inspected W being transported, and a control unit 6 that has a determination unit 6c that performs image processing on the inspection images of the item to be inspected W stored in the image storage unit 4 and determines the quality state of the item to be inspected W from the image processing results. An image processing unit 8 that can be powered by a power supply unit 7 is connected to the connection interface 6a of the control unit 6, and the control unit 6 causes the image processing unit 8 to perform at least a part of the image processing.

Description

Technical Field

[0005]

[0001] The present invention relates to an article inspection apparatus that images an article to be inspected that is conveyed and inspects the quality of the article by performing image processing on the inspection image of the article obtained by the imaging.

Background Art

[0002] In recent years, in the inspection function in image processing, for example, as disclosed in Patent Documents 1 to 3 below, higher-performance inspection using AI (Artificial Intelligence) has come to be used.

[0003] Since AI has a larger amount of calculation than image processing based on conventional rule-based processing, in inspections that require a short processing time, for example, as disclosed in Patent Document 3 below, instead of a CPU (Central Processing Unit), a GPU (Graphics Processing Unit) with high arithmetic processing ability may be used.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, while GPUs are capable of high-speed processing, they have problems such as high power consumption, high heat generation, a high failure rate due to these factors, and large enclosure size. In addition, they require a dedicated power supply, and in environments where waterproofing and dust control are necessary, such as food manufacturing lines, they often have a sealed structure, which increases the cost of creating an adequate cooling environment.

[0006] Therefore, the present invention has been made in view of the above problems, and aims to provide an article inspection device that can perform efficient image processing without requiring an additional power supply. [Means for solving the problem]

[0007] To achieve the above objective, the article inspection apparatus described in claim 1 of the present invention includes an image storage unit 4 that stores inspection images obtained by imaging an article being transported, An article inspection apparatus 1 comprises a control unit 6 having a determination unit 6c that performs image processing on the inspection image stored in the image storage unit and determines the quality state of the article from the image processing result, The image processing unit 8 is connected via a predetermined interface capable of supplying power, and the control unit causes the image processing unit to perform at least a portion of the image processing.

[0008] The article inspection apparatus described in claim 2 of the present invention is an article inspection apparatus according to claim 1, The aforementioned image processing includes image processing to which a trained model is applied. The control unit 6 is characterized by causing the image processing unit 8 to perform image processing to which the trained model is applied.

[0009] The article inspection apparatus described in claim 3 of the present invention is an article inspection apparatus according to claim 1, The aforementioned image processing includes rule-based image processing that applies a predetermined image processing algorithm, The control unit 6 is characterized by causing the image processing unit 8 to perform the rule-based image processing.

[0010] The article inspection apparatus described in claim 4 of the present invention is an article inspection apparatus according to claim 1, The image processing unit 8 comprises a first image processing unit 8A which includes image processing to which a trained model is applied as the image processing, and a second image processing unit 8B which includes rule-based image processing to which a predetermined image processing algorithm is applied as the image processing. The control unit 6 is characterized by causing the first image processing unit to perform image processing to which the trained model is applied, and causing the second image processing unit to perform rule-based image processing.

[0011] The article inspection device described in claim 5 of the present invention is an article inspection device according to any one of claims 1 to 4, The control unit 6 is characterized by cutting off the power supply to the image processing unit 8 that is not used for inspecting the quality status of the article.

[0012] The article inspection apparatus described in claim 6 of the present invention is an article inspection apparatus according to claim 2 or 4, The aforementioned trained model is characterized by being one of the following: foreign matter, missing parts, or shape defects. [Effects of the Invention]

[0013] According to the present invention, image processing can be efficiently performed according to inspection conditions without requiring an additional power supply. [Brief explanation of the drawing]

[0014] [Figure 1] This is a block diagram showing the schematic configuration of a first embodiment of the article inspection device according to the present invention. [Figure 2] This is a block diagram showing a schematic configuration of a second embodiment of the article inspection device according to the present invention. [Figure 3] This is a block diagram showing a schematic configuration of a third embodiment of the article inspection device according to the present invention. [Modes for carrying out the invention]

[0015] Hereinafter, embodiments for implementing the present invention will be described in detail with reference to the accompanying drawings.

[0016] [First Embodiment] The first embodiment of the article inspection apparatus according to the present invention will be described with reference to FIG. 1. As shown in FIG. 1, the article inspection apparatus 1(1A) of the first embodiment images a conveyed inspection object W (article to be inspected), and performs image processing on the inspection image of the inspection object W obtained by the imaging to inspect the quality (foreign matter, defect, etc.) of the inspection object W. It is schematically configured including a conveyance unit 2, an imaging unit 3, an image storage unit 4, a display operation unit 5, a control unit 6, a power supply unit 7, and an image processing unit 8.

[0017] The conveyance unit 2 sequentially conveys the inspection object W at predetermined intervals with respect to the conveyance direction A. For example, a loop-shaped conveyance belt 2a is wound around a plurality of conveyance rollers 2b, and a conveyor that can sequentially convey the inspection object W in the conveyance direction A (right direction in FIG. 1) by the upper running section 2c of the conveyance belt 2a is supported by a housing not shown. The conveyance roller 2b is rotationally driven by a motor not shown and is controlled by the control unit 6 so as to obtain a predetermined conveyance speed.

[0018] <� The imaging unit 3 acquires an inspection image of the inspection object W conveyed by the conveyance unit 2, and includes an X-ray generator 3a disposed at a predetermined height and separated above an inspection space not shown in the middle of the conveyance unit 2, and an X-ray detector 3b disposed opposite to the X-ray generator 3a within the conveyance unit 2.

[0019] The X-ray generator 3a generates X-rays having a wavelength and intensity corresponding to its tube current and tube voltage by a known X-ray tube, and can irradiate the inspection object W and samples (good products of the inspection object W, foreign matters, etc.) on the conveyance belt 2a with fan beam-shaped X-rays orthogonal to the conveyance direction A of the conveyance unit 2 through the X-ray window portion of an outer casing not shown.

[0020] The tube current and tube voltage of the X-ray tube are preferably adjusted according to the material and size of the object W under inspection (especially the dimensions in the direction of X-ray transmission). For new varieties, the settings are determined or selected by test imaging using the object W under inspection or a sample to obtain appropriate contrast.

[0021] The X-ray detector 3b includes an X-ray line sensor, which comprises a scintillator and a photodiode array (not shown). The scintillator converts X-rays into light, and the photodiode array converts this light into an electrical signal to output an X-ray image based on X-ray transmission data.

[0022] The image storage unit 4 temporarily stores the X-ray image obtained from the X-ray transmission data received from the X-ray detector 3b of the imaging unit 3 as the inspection image of the object W under inspection.

[0023] The display operation unit 5 is operated to switch the power, transport unit 2, and imaging unit 3 ON / OFF, and to perform various settings and displays necessary when inspecting the quality of the object W under inspection.

[0024] The control unit 6, including the image processing unit 8, provides overall control for all parts of the item inspection device 1 (transport unit 2, imaging unit 3, image storage unit 4, and display operation unit 5), and is configured to include a connection interface 6a, an image processing unit 6b, and a determination unit 6c.

[0025] The control unit 6 is connected to the power supply unit 7, and the power supply unit 7 supplies the basic power necessary to drive each part of the item inspection apparatus 1, including the image processing unit 8 (transport unit 2, imaging unit 3, image storage unit 4, display operation unit 5, and control unit 6).

[0026] The control unit 6 controls the image processing unit 8 to perform predetermined image processing. This allows the image processing unit 8 to perform a portion of the image processing that the control unit 6 (CPU) should perform. Furthermore, if there are multiple image processing units 8, the image processing can be distributed and performed by each of the image processing units 8.

[0027] The control unit 8 can also control the image processing unit 8 to perform image processing that applies a trained model or image processing based on registered rules (image processing that applies a predetermined image processing algorithm: rule-based image processing, described later) as predetermined image processing. This allows the image processing unit 8 to perform image processing that applies a trained model or image processing based on registered rules. Furthermore, if the image processing unit 8 consists of multiple units, it is possible to have multiple image processing units 8 perform only image processing that applies a trained model, or to have multiple image processing units 8 perform only image processing based on registered rules, or to appropriately distribute the image processing that applies a trained model and the image processing based on registered rules among multiple image processing units 8.

[0028] The connection interface 6a is connected to the image storage unit 4 and the determination unit 6c, and is also connected to the image processing unit 8 via a detachable terminal (e.g., M2 slot, PCI slot, USB terminal, etc.) on the motherboard of the main body of the item inspection device 1. The interface 6a inputs the inspection image of the object to be inspected W stored in the image storage unit 4 to the image processing unit 8, and inputs the image data after image processing from the image processing unit 8 to the determination unit 6c.

[0029] The image processing unit 6b sequentially performs image processing on the inspection image of the object to be inspected W input from the image storage unit 4 according to a predetermined image processing algorithm. More specifically, the image processing performed by the image processing unit 6b is rule-based image processing (hereinafter referred to as rule-based image processing) that does not use a trained model. It performs image processing using a predetermined image processing algorithm that shows the procedure for performing image processing such as analysis processing using the difference in values ​​with the surroundings, frequency analysis, and pattern matching on the input inspection image, and outputs image data of the image processing result for making a judgment.

[0030] The determination unit 6c performs quality state determination processing (determination of the presence or absence of abnormalities...for example, determination of the presence or absence of foreign matter, determination of the presence or absence of missing parts, determination of the presence or absence of shape defects, etc.) based on the image data processed by the image processing unit 6b and the image data from the image processing unit 8. More specifically, in the image data resulting from the rule-based image processing, the determination unit 6c determines that there is a defective part if, for example, there is a pixel (or cluster of pixels) whose gradation (=defect level) is above the threshold for defect determination. Also, in the image data resulting from the AI ​​image processing described later, the determination unit 6c determines that there is a defective part if there is a pixel (or cluster of pixels) whose confidence level is above the threshold. Then, based on the determination result, the determination unit 6c outputs an OK signal (a signal indicating that there is no abnormality in the object W under inspection) or an NG signal (a signal indicating that there is an abnormality in the object W under inspection).

[0031] The power supply unit 7 is connected to the connection interface 6a and, under the control of the control unit 6, supplies the basic power necessary to drive each part of the item inspection apparatus 1 (transport unit 2, imaging unit 3, image storage unit 4, display operation unit 5, and control unit 6), including the image processing unit 8. At the same time, the control unit 6 disconnects the power supply to the image processing unit 8, which is not used for inspecting the quality status of each type of item W under inspection.

[0032] The image processing unit 8 is connected to the connection interface 6a via a detachable terminal (e.g., M2 slot, PCI slot, USB terminal, etc.) on the motherboard of the main body of the item inspection device 1.

[0033] The image processing unit 8, under the control of the control unit 6, can run a binary image processing program and performs image processing using a dedicated computing unit that does not include a GPU, in order to extract (or make easier to obtain) predetermined information from the inspection image of the object W under inspection.

[0034] The image processing unit 8 is preferably composed of an FPGA (Field Programmable Gate Array) unit or an image processing component (VPU: Vision Processing Unit) that has lower power consumption, heat generation, and enclosure size compared to a GPU, as it does not require a dedicated power supply like a GPU.

[0035] In Figure 1, the image processing unit 8 consists of two image processing units 8A, each equipped with a trained model 8a and an image processing unit 8b.

[0036] The trained model 8a consists of a trained model that has been trained on, for example, defective products (e.g., foreign objects, missing parts, shape defects, etc.) for each type of product W under inspection. Note that the trained model 8a is not limited to a trained model that has been trained on defective products for each type of product W under inspection; it may also be a trained model that has been trained on OK products for each type of product W under inspection, or a trained model that has been trained on both OK and defective products for each type of product W under inspection.

[0037] The image processing unit 8b performs image processing necessary to determine whether or not there is an abnormality in the object W under inspection by inference based on the trained model 8a, as an image processing operation to which the trained model 8a is applied. More specifically, the image processing performed by the image processing unit 8b using the trained model 8a (hereinafter referred to as AI image processing) is image processing that uses the trained model 8a, which is a model obtained as a result of machine learning performed in advance to take information based on the inspection image of the object W under inspection as input and output information that can derive multiple inference values.

[0038] AI image processing uses deep learning-based classification or anomaly detection functions based on the inspection image of the object W under inspection, or the inspection image of the object W after predetermined filtering, to determine whether the quality state of the object W under inspection is normal or abnormal, using a confidence level (inferred value of the processing result using the trained model 8a).

[0039] Here, classification is the process of performing image class classification that can identify the type of object W under inspection by extracting features from the input image and learning the decision boundary, while anomaly detection is the process of detecting abnormal parts in the input image, such as partial defects in the contents of the object W under inspection or irregularities that fall outside the normal range.

[0040] In the example shown in Figure 1, two image processing units 8A are depicted as the image processing unit 8, but the number is not limited. The configuration only requires at least one image processing unit 8A, and the necessary number can be connected to the connection interface 6a depending on the processing content.

[0041] Next, we will describe the operation when the quality of an object to be inspected W is inspected using the article inspection device 1A of the first embodiment configured as described above.

[0042] First, when the transport unit 5 is operated to drive the transport unit 2 and the imaging unit 3 (X-ray generator 3a, X-ray detector 3b), the object to be inspected W is sequentially transported by the transport unit 2 in the transport direction A at predetermined intervals.

[0043] Then, X-rays are irradiated from the X-ray generator 3a onto the object W to be inspected on the conveyor belt 2a, and the X-rays that pass through the object W are detected by the X-ray detector 3b. An X-ray image is then output based on X-ray transmission data corresponding to the amount of X-rays transmitted.

[0044] Then, the X-ray images obtained from the X-ray transmission data from the X-ray detector 3b are sequentially stored in the image storage unit 4 as inspection images of the object W under inspection. Subsequently, the image processing unit 6b of the control unit 6 performs image processing sequentially on the inspection images of the object W under inspection input from the image storage unit 4 according to a predetermined image processing algorithm.

[0045] Furthermore, the image processing unit 8 (8A) is powered by the power supply unit 7 via the connection interface 6a under the control of the control unit 6, and performs AI image processing by applying the trained model 8a. For example, if the image processing unit 8 is composed of two image processing units 8A, one image processing unit 8A can perform AI image processing by applying the trained model 8a for defective products of product type α with foreign matter, and the other image processing unit 8A can perform AI image processing by applying the trained model 8a for defective products of product type α with shape defects. In this case, if inspection for shape defects in product type α is not required, the control unit 6 will cut off the power supply to the other image processing unit 8A because it is no longer needed for inspecting the quality state of the inspected object W.

[0046] Then, the determination unit 6c of the control unit 6 performs quality status determination processing (for example, determination processing for foreign matter, missing parts, shape defects, etc.) based on the image data processed by the image processing unit 6b and the image data from the image processing unit 8 (8A). If there is no abnormality in the object W to be inspected, it outputs an OK signal, and if there is an abnormality in the object W to be inspected, it outputs an NG signal.

[0047] Furthermore, the display operation unit 5 displays the quality status of the inspected object W based on the determination result of the determination unit 6c of the control unit 6 (display of the inspection image of the inspected object W including contents and foreign matter, display of a projection image corresponding to the inspection image of the inspected object W, production capacity, total number of inspected objects W, total number of OKs relative to the total number of inspections, total number of NGs, total number of NGs with foreign matter, etc.).

[0048] [Second Embodiment] Next, a second embodiment of the article inspection device according to the present invention will be described with reference to Figure 2. In Figure 2, the components other than the image processing unit 8 are the same as those of the article inspection device 1 of the first embodiment in Figure 1, so the same numbers are used for the same components as in the article inspection device 1 of the first embodiment, and their descriptions are omitted.

[0049] As shown in Figure 2, the image processing unit 8 in the article inspection device 1(1B) of the second embodiment is composed of two types of image processing units 8A and 8B.

[0050] The two types of image processing units 8A and 8B are preferably composed of image processing components (VPU: Vision Processing Unit) that have lower power consumption, heat generation, and enclosure size compared to FPGA (Field Programmable Gate Array) units and GPUs, similar to the first embodiment.

[0051] In the example shown in Figure 2, the image processing unit 8 is illustrated with one image processing unit 8A and one image processing unit 8B, but the number is not limited. The configuration only requires at least one of each of the two types of image processing units 8A and 8B, and the necessary number can be connected to the connection interface 6a depending on the processing content.

[0052] The image processing unit 8A includes a trained model 8a and an image processing unit 8b with the same configuration as in the first embodiment. The processing contents of the trained model 8a and the image processing unit 8b are the same as in the first embodiment, so their description will be omitted.

[0053] The image processing unit 8B includes an image processing unit 8c that performs a portion of the rule-based image processing carried out by the image processing unit 6b of the control unit 6. The image processing unit 8c performs rule-based image processing in parallel when the image processing unit 6b of the control unit 6 alone cannot keep up with the processing time, such as when processing multiple inspection items with multiple image processing algorithms or when using complex image processing algorithms with short inspection intervals.

[0054] For example, some of the multiple image processing algorithms can be executed by the image processing unit 8B's image processing unit 8c, and the image processing unit 6b of the control unit 6 and the image processing unit 8c of the image processing unit 8B can be processed in parallel. Alternatively, a complex image processing algorithm can be executed by switching between the image processing unit 6b of the control unit 6 and the image processing unit 8B's image processing unit 8c in the order of inspection, thereby overlapping the execution of the same image processing algorithm. This makes it possible to inspect each object W even if the objects W are transported at short intervals.

[0055] [Third Embodiment] Next, a third embodiment of the article inspection device according to the present invention will be described with reference to Figure 3. In Figure 3, the components other than the image processing unit 8 are the same as those of the article inspection device 1 of the first embodiment in Figure 1, so the same numbers are used for the same components as in the article inspection device 1 of the first embodiment, and their descriptions are omitted.

[0056] As shown in Figure 3, the image processing unit 8 in the third embodiment of the article inspection apparatus 1(1C) is composed of two image processing units 8B.

[0057] The image processing unit 8B is preferably composed of an image processing component (VPU: Vision Processing Unit) that has lower power consumption, heat generation, and enclosure size compared to an FPGA (Field Programmable Gate Array) unit or a GPU, similar to the first and second embodiments.

[0058] In the example shown in Figure 3, two image processing units 8B are depicted as the image processing unit 8, but the number is not limited. The configuration only requires at least one image processing unit 8B, and as many as necessary can be connected to the connection interface 6a depending on the processing content.

[0059] Similar to the second embodiment, the image processing unit 8B includes two image processing units 8c that perform a portion of the rule-based image processing carried out by the image processing unit 6b of the control unit 6. Each image processing unit 8c performs rule-based image processing in parallel when the image processing unit 6b of the control unit 6 alone cannot keep up with the processing time, such as when processing multiple inspection items with multiple image processing algorithms or when using complex image processing algorithms with short inspection intervals.

[0060] For example, some of the multiple image processing algorithms can be implemented by the image processing units 8c of each of the two image processing units 8B, and the image processing units 6b of the control unit 6 and 8c of the image processing units 8B can be processed in parallel. Alternatively, a complex image processing algorithm can be implemented by switching between the image processing units 6b of the control unit 6 and 8c of the two image processing units 8B in the order of inspection, thereby overlapping the implementation of the same image processing algorithm. This makes it possible to inspect each object W even if it is transported at short intervals.

[0061] By the way, in the embodiment described above, we explained an example in which the imaging unit 3 is equipped with an X-ray generator 3a and an X-ray detector 3b, and while transporting the object to be inspected W, X-rays are irradiated from the X-ray generator 3a, and the X-rays that have passed through the object to be inspected W are detected by the X-ray detector 3b, and the X-ray image obtained is used as an inspection image to inspect the quality of the object to be inspected W. However, the invention is not limited to this, and the imaging unit 3 can be configured in any way that can obtain an inspection image of the object to be inspected W.

[0062] Specifically, the imaging unit 3 is equipped with a light emitter and a light receiver, and instead of X-rays as electromagnetic waves, it emits light from the light emitter onto the object to be inspected W, and the light that has passed through or been projected onto the object to be inspected W is detected by the light receiver to inspect the quality of the object to be inspected W. It can also be applied to an object inspection device in which the imaging unit 3 is equipped with a light source and a camera, and the light source and camera are positioned opposite each other with a gap S between the upstream conveyor 4 and the downstream conveyor 5, and the light source is emitted onto the object to be inspected W, and the quality of the object to be inspected is inspected from the image captured by the camera.

[0063] Furthermore, in the above-described embodiment, the determination unit 6c determines the quality state of the object to be inspected W based on the image data from the image processing unit 6b and the image data from the image processing unit 8. However, the image processing unit 6b and the image processing unit 8 may determine whether there are any abnormalities in the corresponding object to be inspected W (such as foreign objects, missing parts, or shape defects), and the determination unit 6c may make a comprehensive determination based on the respective determination results using conditions such as AND.

[0064] Thus, according to this embodiment, an image processing unit 8 that performs image processing using a dedicated computing unit without a GPU, which is detachable from the connection interface 6a of the control unit 6 of the item inspection device 1 (1A, 1B, 1C), is used, and the power supply unit 7 connected via the connection interface 6a of the control unit 6 is also used as the power supply for this image processing unit 8. As a result, an item inspection device can be provided that does not require an additional power supply and can efficiently perform image processing according to inspection conditions.

[0065] Furthermore, by using the highly expandable image processing unit 8, it becomes possible to provide a device configuration that is sufficient for a variety of functions.

[0066] To further explain, in this embodiment, instead of a GPU, a VPU is used as a detachable image processing unit 8 attached to the connection interface 6a of the control unit 6 of the item inspection device 1 (1A, 1B, 1C). This VPU has lower power consumption, heat generation, and housing size compared to a GPU, thereby enabling the installation of an AI inference module within the item inspection device.

[0067] Here, while the VPU as an image processing unit 8 has inferior processing power compared to a GPU, resulting in disadvantages such as slower processing speed and the inability to perform parallel processing on a single unit, the control unit 6a of the item inspection device 1 employs a mechanism that allows multiple VPU image processing units 8 to be attached and detached depending on the application. This makes it possible to provide a sufficiently functional device configuration by adjusting the number of VPU image processing units 8 installed to meet customer needs, from customers who only need one VPU image processing unit 8 to customers who use multiple VPU image processing units 8 to perform processing-intensive functions or perform various functions simultaneously.

[0068] In addition, the VPU's image processing unit 8 does not require a dedicated power supply like a GPU, and can use the power supply unit 7 attached to the item inspection device 1. Furthermore, because it generates less heat than a GPU, improvements can be expected in the failure rate of the VPU itself and surrounding electronic equipment.

[0069] Furthermore, for compact item inspection devices intended for installation in locations with limited space, the aforementioned low power consumption, low heat generation, and small enclosure size, along with minimal impact on the device, make it easier to miniaturize the device compared to GPUs.

[0070] Furthermore, when it is necessary to perform learning on captured images in real time, the VPU's image processing unit 8 can be used to perform learning faster than with the CPU. In addition to performing learning even when the production line is not running, by increasing the number of VPU image processing units 8 installed, it becomes possible to perform background learning while inspections are being carried out on the production line.

[0071] Although the image processing unit 8 of the VPU was explained using a Vision Processing Unit as an example, it could also be an image processing unit for video processing (Video Processing Unit) or a Visual Processing Unit (Visual Processing Unit).

[0072] The best mode of the article inspection apparatus according to the present invention has been described above, but the present invention is not limited by this description and drawings. That is, other modes, examples, and operational techniques based on this mode, as made by those skilled in the art, are all included in the scope of the present invention. [Explanation of symbols]

[0073] 1 (1A, 1B, 1C) Item inspection device 2. Conveying section 2a Conveyor belt 2b Conveyor roller 2c Upper running section 3. Imaging Unit 3a X-ray generator 3b X-ray detector 4 Image storage unit 5 Display operation section 6 Control Unit 6a Connection Interface 6b Image Processing Unit 6c Judgment section 7 Power supply unit 8(8A,8B) Image Processing Unit 8a Pre-trained model 8b,8c Image Processing Unit W: Object under inspection A Conveying direction

Claims

1. An image storage unit (4) stores inspection images obtained by imaging the transported items, In an article inspection device (1), the device comprises a control unit (6) having a determination unit (6c) that performs image processing on the inspection image stored in the image storage unit and determines the quality state of the article from the image processing result, An article inspection apparatus characterized in that an image processing unit (8) is connected via a predetermined interface capable of supplying power, and the control unit causes the image processing unit to perform at least a part of the image processing.

2. The aforementioned image processing includes image processing to which a trained model is applied. The article inspection apparatus according to claim 1, characterized in that the control unit (6) causes the image processing unit (8) to perform image processing to which the learned model is applied.

3. The aforementioned image processing includes rule-based image processing that applies a predetermined image processing algorithm, The article inspection apparatus according to claim 1, characterized in that the control unit (6) causes the image processing unit (8) to perform the rule-based image processing.

4. The image processing unit (8) comprises a first image processing unit (8A) which includes image processing to which a trained model is applied as the image processing, and a second image processing unit (8B) which includes rule-based image processing to which a predetermined image processing algorithm is applied as the image processing. The article inspection apparatus according to claim 1, characterized in that the control unit (6) causes the first image processing unit to perform image processing to which the trained model is applied, and causes the second image processing unit to perform rule-based image processing.

5. The article inspection apparatus according to any one of claims 1 to 4, characterized in that the control unit (6) disconnects the power supply to the image processing unit (8) that is not used for inspecting the quality state of the article.

6. The article inspection apparatus according to claim 2 or 4, characterized in that the learned model is one of foreign matter, missing parts, or shape defects.

Images