Appearance inspection device, image storage method, and image storage program

Abstract

To provide an appearance inspection device, an image saving method, and an image saving program that can save many effective inspection images. [Solution] The visual inspection device 100 comprises a comprehensive general inspection unit 10 that performs a pass / fail judgment on a captured image of a workpiece by integrating general inspections using multiple different types of inspection items and outputs a comprehensive general inspection result indicating pass / fail, an AI inspection unit 20 that performs a pass / fail judgment based on the captured image using a trained machine learning model that can learn and outputs an AI inspection result indicating pass / fail, and an image storage unit 30 that can store the captured image. The AI ​​inspection unit 20 assigns an AIOK flag to the captured image if the image satisfies conditions based on the AI ​​inspection result and the comprehensive general inspection result, and the image storage unit 30 makes the captured image saveable if the AIOK flag has been assigned to the captured image.

Description

[Technical Field]

[0001] The present invention relates to an appearance inspection device, an image storage method, and an image storage program. [Background technology]

[0002] Visual inspection devices that perform visual inspections of products have been available for some time. In visual inspections using conventional visual inspection devices, an inspection (inspection 0) is performed to image the workpiece and then detect the workpiece's position, followed by multiple inspections (inspections 1, 2, 3, ...) that are specialized for detecting various possible visual defects. In conventional visual inspections, if none of the inspections result in a defective product (NG), the overall judgment result is that the inspected workpiece is a good product (OK). For example, it has been possible to set the storage conditions for the inspection image (whether to save the inspection image or not) for each overall judgment result (separated into OK and NG cases) (see, for example, Patent Document 1).

[0003] Conventional image storage methods only allowed the storage conditions for inspection images to be set based on the simple condition of the overall visual inspection result, which sometimes made it impossible to secure a sufficient number of valid inspection images for the intended purpose. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 6999150 Summary of the Invention [Problem to be solved by the invention]

[0005] An object of the present invention is to provide an appearance inspection apparatus, an image saving method, and an image saving program that can save many effective inspection images. [Means for solving the problem]

[0006] (1) An appearance inspection device according to one embodiment of the present invention comprises a comprehensive general inspection unit that performs a pass / fail judgment on a captured image of a workpiece by integrating general inspections using multiple different types of inspection items and outputs a comprehensive general inspection result indicating pass / fail; an AI inspection unit that performs a pass / fail judgment based on the captured image using a trained machine learning model that can learn and outputs an AI inspection result indicating pass / fail; and an image storage unit that can store the captured image, wherein the AI ​​inspection unit assigns an AIOK flag to the captured image if the image satisfies conditions based on the AI ​​inspection result and the comprehensive general inspection result, and the image storage unit enables the captured image to be stored if the AIOK flag has been assigned to the captured image. (2) In the above (1), the image saving unit may save the captured image when an image saving setting is ON. (3) In (1) or (2) above, an inspection condition determination unit may be provided that outputs a determination result as to whether the average density of the inspection execution area in the captured image is within a preset range, and enables the AI ​​inspection unit to perform an operation if the determination result is within the set range. (4) An image storage method according to one embodiment of the present invention includes a comprehensive general inspection step of making a pass / fail judgment on a captured image of a workpiece by integrating general inspections using multiple different types of inspection items and outputting a comprehensive general inspection result indicating pass / fail; an AI inspection step of making a pass / fail judgment based on the captured image using a trained machine learning model that can be trained and outputting an AI inspection result indicating pass / fail; and an AIOK flag determination step of determining whether the captured image has an AIOK flag if the captured image satisfies predetermined image storage conditions, and making the captured image saveable if the captured image has the AIOK flag. (5) In the above (4), if the AIOK flag determination step is NO, a first overall result determination step is executed to determine whether the overall general inspection result is OK, and if the overall general inspection result is OK, the captured image may be made saveable. (6) In the above (5), if the first overall result determination step results in a NO judgment, a second overall result determination step is executed to determine whether the overall general test result is NG or not, and if the second overall result determination step results in a YES judgment, the captured image may be made saveable. (7) In any of (4) to (6) above, an image storage condition determination step may be included in which a determination result as to whether the average density of the inspection execution area in the captured image is within a preset range is output, and if the determination result in the image storage condition determination step is within the set range, the AI ​​inspection step may be executed. (8) An image saving program according to one embodiment of the present invention causes a computer to execute a comprehensive general inspection function that performs a pass / fail judgment on a captured image of a workpiece by integrating general inspections using multiple different types of inspection items and outputting a comprehensive general inspection result indicating pass / fail; an AI inspection function that performs a pass / fail judgment based on the captured image using a trained machine learning model that can learn, and outputs an AI inspection result indicating pass / fail; a flag assignment function that assigns an AIOK flag to the captured image if the captured image meets the predetermined image saving conditions; and an AIOK flag determination function that determines whether the captured image has an AIOK flag if the captured image meets the image saving conditions, and allows the captured image to be saved if the AIOK flag determination function judges YES. (9) In the above (8), if the AIOK flag determination function judges NO, a first overall result determination function is executed to determine whether the overall general test result is OK or not, and if the first overall result determination function judges YES, the captured image may be made saveable. (10) In the above (9), if the first overall result determination function judges NO, a second overall result determination function is executed to determine whether the overall general test result is NG or not, and if the second overall result determination function judges YES, the captured image may be made saveable. (11) In any of (8) to (10) above, an image storage condition determination function may be executed to output a determination result as to whether the average density of the inspection execution area in the captured image is within a preset range, and if the image storage condition determination function determines that the determination result is within the set range, the AI ​​inspection function may be executed. [Effects of the Invention]

[0007] According to the present invention, it is possible to provide a visual inspection device, an image saving method, and an image saving program that can save many effective inspection images. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a functional block diagram illustrating an overview of a visual inspection apparatus according to an embodiment. [Figure 2] FIG. 1 is a flow diagram of an image saving method according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] (First embodiment) A first embodiment of the present invention will be described in detail below with reference to the drawings. Fig. 1 is a functional block diagram illustrating an overview of an appearance inspection apparatus 100 according to an embodiment. Fig. 2 is a flow diagram of an image saving method according to an embodiment. Note that, hereinafter, parts having common functions may be assigned the same reference numerals or symbols.

[0010] When AI inspection (which uses a trained machine learning model that can learn from input captured images to infer whether an image is good or bad) is introduced into the visual inspection of workpieces (products), it is conceivable that the stored images that have been judged as good by the AI ​​inspection will be visually checked to ensure that no defects have been leaked (that is, actual defective products are judged as good).However, image processing devices have a limit to their processing power (the number of images that can be saved at one time), so it is necessary to selectively save valid images that suit the purpose.

[0011] Therefore, based on the captured image P obtained by capturing an image of the workpiece, the visual inspection device 100, which includes general inspection (non-AI inspection) and AI inspection, determines whether the workpiece is good (OK) or bad (NG), or AIOK (an evaluation where the AI ​​inspection result is OK). Here, if the visual inspection device 100 determines that the workpiece is AIOK, it is possible to save the captured image P according to preset conditions for saving the captured image P. This makes it possible to selectively save effective images that match the purpose by changing the conditions depending on the purpose (collection of captured images as evidence, learning or training of a machine learning model M used in AI inspection, etc.). Therefore, it is possible to provide a visual inspection device 100 that can save many effective inspection images.

[0012] (visual inspection equipment) The appearance inspection apparatus 100 according to the embodiment can be used to inspect the appearance of products such as capacitors and other electronic components. The appearance inspection apparatus 100 can be used to sort products into good products (also called OK products) and bad products (also called NG products) based on their appearance. Unless otherwise specified below, the subject of information processing is the appearance inspection apparatus 100. The appearance inspection apparatus 100 can realize each function related to the image saving method by having the CPU of a computer process information using an image saving program stored in a storage medium (memory).

[0013] As shown in Figure 1, the visual inspection device 100 includes a comprehensive general inspection unit 10 that performs a pass / fail judgment on a captured image P of a workpiece based on a comprehensive general inspection of multiple different types of inspection items and outputs a comprehensive general inspection result 10R indicating pass / fail, an AI inspection unit 20 that performs a pass / fail judgment based on the captured image P using a trained machine learning model M that can be trained and outputs an AI inspection result 20R indicating pass / fail, and an image storage unit 30 that can store the captured image P.

[0014] General testing here refers to non-AI testing that is different from the AI ​​testing described below.

[0015] The overall general inspection unit 10 performs a pass / fail judgment on the captured image P of the workpiece by integrating general inspections based on a plurality of different types of inspection items, and outputs an overall general inspection result 10R indicating pass or fail.

[0016] The AI ​​inspection unit 20 is equipped with a trained machine learning model M using a neural network. Even after being trained once, the machine learning model M can adapt to newly input captured images P (data), allowing for additional learning through fine tuning, transfer learning, etc. The AI ​​inspection unit 20 uses the machine learning model M to analyze the features of the captured images P, automatically judge whether the images are good or bad based on previously learned data on good and bad products, and output an AI inspection result 20R indicating whether the images are good or bad. The machine learning model M may be a standalone model or may be configured by combining multiple types of models or algorithms. The AI ​​inspection unit 20 may be independent and not be affected by the results of the comprehensive general inspection unit 10.

[0017] The AI ​​inspection unit 20 assigns an AIOK flag to a captured image P if it satisfies conditions based on the AI ​​inspection result 20R and the overall general inspection result 10R. This allows the captured image P to be extracted for storage, taking into account both the AI ​​inspection result 20R and the overall general inspection result 10R. Specifically, the AI ​​inspection unit 20 assigns an AIOK flag to a captured image P if the AI ​​inspection result 20R contains one or more pass judgments, there are no fail judgments, and there are no fail judgments in the overall general inspection result 10R. Here, by setting the condition "there are one or more pass judgments in the AI ​​inspection result 20R," it is possible to prevent the AIOK flag from being set even if none of the multiple prepared AI inspections are performed. Furthermore, by setting the condition "there are no fail judgments in the AI ​​inspection result 20R," it is possible to prevent the AIOK flag from being set if the result of one AI inspection is OK, even if some of the prepared AI inspections contain a fail judgment.

[0018] The image storage unit 30 can store the captured image P. The image storage unit 30 can store the captured image P as an image file including input image data, processed image data, and metadata. The image storage unit 30 may have a function to store the captured image P in a storage medium such as a hard disk drive (HDD), a solid-state drive (SSD), or a flash memory. The image storage unit 30 may be implemented in a computer system including a storage medium, a CPU, storage, and an input / output device. The image storage unit 30 may include a data verification function using a checksum or hash value to ensure data integrity, as appropriate. The image storage unit 30 may be configured to cooperate with a database management system (DBMS) for searching and managing the stored captured image P. The image storage unit 30 may cooperate with an image processing module as appropriate to perform format conversion, compression processing, and tagging before storage. The image storage unit 30 may have a function to store images in different formats, such as JPEG, PNG, and TIFF.

[0019] Here, the image storage unit 30 is able to store a captured image P when the AI ​​inspection unit 20 assigns an AIOK flag to the captured image P (when the following conditions are met: there is one or more pass judgments in the AI ​​inspection results 20R, there are no fail judgments in the AI ​​inspection results 20R, and there are no fail judgments in the overall general inspection results 10R), i.e., when an AIOK flag is assigned to the captured image P. This allows selective storage of valid images that are in line with the purpose, taking into account not only captured images P that have been judged pass in the overall general inspection results 10R, but also the AI ​​inspection results 20R, thereby improving storage usage efficiency. Therefore, a visual inspection device that can store many valid inspection images can be provided.

[0020] The above-mentioned condition for the AI ​​inspection unit 20 to assign an AIOK flag to a captured image P, "the overall general inspection result 10R is not a NG judgment," may be replaced with "there is no NG judgment in other inspections not connected to the AI ​​inspection." Here, "other inspections not connected to the AI ​​inspection" refers to inspections other than the "optional inspection." When using AI inspection, the user can perform AI inspection on images that have been judged NG in the "optional inspection."

[0021] Specifically, before the AIOK flag determination step S1 described later, the visual inspection device 100 appropriately determines whether the captured image P satisfies the conditions for saving the image (image saving condition determination step S4). That is, the visual inspection device 100 checks whether the average density setting for the inspection area is ON and whether the average density is outside the range. Specifically, if the workpiece is out of a specific inspection area within the camera's field of view and the captured image P was not captured correctly, the image storage unit 30 does not store the captured image P because the average density of the captured image P is outside the preset range. Conversely, if the workpiece is located within a predetermined inspection area within the camera's field of view and the captured image P was captured correctly, the image storage unit 30 stores the captured image P because the average density of the captured image P is within the preset range. That is, the image storage unit 30 determines whether the captured image P is located within the predetermined inspection area and has been captured correctly based on the average density of the captured image P. The average density can be calculated, for example, by dividing the sum of the luminance values ​​of each pixel in the captured image P by the number of pixels. This makes it possible to identify captured images P that clearly do not need to be saved due to an imaging error or the like while reducing the computational load. Also, it is possible to prevent the image saving unit 30 from saving captured images P that have not been captured correctly, such as captured images P that have been captured when the workpiece has shifted from a specific inspection execution area within the camera's angle of view, or when the workpiece is not even within the camera's angle of view.

[0022] When the image saving setting is ON, the image saving section 30 may save the captured image P. In detail, the image saving section 30 is configured to allow a setting in advance as to whether or not to save the captured image P. This allows the user to select whether or not to save the captured image P according to the setting (condition).

[0023] The visual inspection device 100 may include an inspection condition determination unit 40 that outputs a determination result 40R as to whether the average density of the inspection execution area in the captured image P is within a preset range, and enables operation by the AI ​​inspection unit 20 if the determination result 40R is within the set range. This makes it possible to prevent the storage of unnecessary captured images P, such as captured images P in which the workpiece to be inspected is not properly or clearly visible in the inspection execution area of ​​the captured image P.

[0024] (Image saving method) Next, an image saving method will be described using the visual inspection apparatus 100 according to this embodiment. Fig. 2 is a flow diagram of the image saving method. The functions corresponding to the steps of the image saving method described below can be executed by a computer using an image saving program.

[0025] (1) The comprehensive general inspection unit 10 of the visual inspection device 100 performs a pass / fail judgment on the captured image P of the workpiece by integrating general inspections using multiple different types of inspection items, and outputs a comprehensive general inspection result 10R indicating pass or fail (comprehensive general inspection step).

[0026] (2) The AI ​​inspection unit 20 of the visual inspection device 100 judges whether the captured image P is good or bad using a trained machine learning model M that can be trained, and outputs an AI inspection result 20R indicating good or bad (AI inspection step).

[0027] (3) As shown in FIG. 2, it is determined as appropriate whether the captured image P satisfies the image storage conditions (image storage condition determination step S4). Here, if the captured image P does not satisfy the image storage conditions (NO determination in image storage condition determination step S4), the captured image P is not saved and the process ends (END). This makes it possible to identify captured images P that clearly do not need to be saved due to an imaging error or the like while reducing the computational load. Also, it is possible to prevent the image saving unit 30 from saving captured images P that were not captured correctly, such as captured images P captured when the workpiece is shifted from a specific inspection execution area within the camera's angle of view.

[0028] (3-1) More specifically, in the image saving condition determination step S4, a determination result is output as to whether or not the average density of the inspection execution area in the captured image P is within a preset range. Then, if the determination result in the image saving condition determination step S4 is within the set range, the AIOK flag determination step S1 is executed.

[0029] (4) If the captured image P satisfies the image storage conditions (YES in image storage condition determination step S4), an AIOK flag is assigned to the captured image P. In this case, if the captured image P satisfies the image storage conditions, and satisfies the conditions based on the AI ​​inspection result 20R and the overall general inspection result 10R, an AIOK flag may be assigned to the captured image P.

[0030] (4-1) If the captured image P has an AIOK flag (YES determination in AIOK flag determination step S1), and the AI ​​good product storage setting is ON, the captured image P is saved and the process ends (END). If the AI ​​non-good product storage setting is OFF, the captured image P is not saved and the process ends (END). In this way, if the captured image P has an AIOK flag, the captured image P is saved regardless of the overall general inspection result 10R, so that many valid images can be saved.

[0031] (4-2) If the captured image P does not have an AIOK flag (NO in the AIOK flag determination step S1), the process proceeds to the first comprehensive result determination step S2.

[0032] (5) If the AIOK flag determination step S1 is NO, it is determined whether the overall general inspection result 10R is OK or not (first overall result determination step S2). Here, if the overall general inspection result 10R is OK (YES determination in first overall result determination step S2), and the OK save setting is ON, the captured image P is saved and the process ends (END). If the OK save setting is OFF, the captured image P is not saved and the process ends (END). On the other hand, if the overall general inspection result 10R is not OK (NO determination in first overall result determination step S2), the process proceeds to second overall result determination step 3.

[0033] (6) If the first overall result determination step S2 is NO, a pass / fail determination is made based on the overall general inspection result 10R (second overall result determination step S3). Here, if the overall general inspection result 10R is NG (NG determination in the second overall result determination step S3), and the NG save setting is ON, the captured image P is saved and the process ends (END). If the NG save setting is OFF, the captured image P is not saved and the process ends (END). On the other hand, if the overall general inspection result 10R is NG (NO determination in the second overall result determination step S3), the captured image P is not saved and the process ends (END).

[0034] In this way, the image storage method according to this embodiment can selectively store effective images that meet the purpose, taking into account not only the captured images P that are judged to be good in the overall general inspection result 10R, but also the AI ​​inspection result 20R, thereby improving storage utilization efficiency. Therefore, an image storage method that can store many effective inspection images can be provided.

[0035] (Image saving program) The image saving program has the function of causing a computer to execute each process (function) performed at each step in the image saving method described above. Specifically, the image saving program causes the computer to execute, in accordance with each step in the image saving method described above, a comprehensive general inspection function that causes a captured image P of a workpiece to be subjected to a general inspection based on multiple different types of inspection items and output a comprehensive general inspection result 10R indicating pass or fail, an AI inspection function that causes a learnable, trained machine learning model M to be used to determine pass or fail based on the captured image P and output an AI inspection result 20R indicating pass or fail, a flag assignment function that assigns an AIOK flag to the captured image P if the captured image P satisfies preset image saving conditions, and an AIOK flag determination function that determines whether the captured image P has an AIOK flag if the captured image P satisfies the image saving conditions, and allows the captured image P to be saved if the AIOK flag determination function determines YES. In this way, the image saving program according to this embodiment can selectively save valid images that meet the purpose, taking into account not only the captured images P that are judged to be good in the overall general inspection result 10R, but also the AI ​​inspection result 20R, thereby improving storage usage efficiency. Therefore, an image saving program that can save many valid inspection images can be provided.

[0036] The technical scope of the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the spirit of the present invention. Furthermore, the components in the above-described embodiments can be appropriately replaced with well-known components without departing from the spirit of the present invention. Furthermore, the above-described modifications can be appropriately combined without departing from the spirit of the present invention.

[0037] As described above, the visual inspection device 100 according to the embodiment includes a comprehensive general inspection unit 10 that performs a pass / fail judgment on a captured image P of a workpiece based on a general inspection of multiple different types of inspection items and outputs a comprehensive general inspection result 10R indicating pass / fail; an AI inspection unit 20 that performs a pass / fail judgment based on the captured image P using a trained machine learning model M that can learn and outputs an AI inspection result 20R indicating pass / fail; and an image storage unit 30 that can store the captured image P. The AI ​​inspection unit 20 assigns an AIOK flag to the captured image P if the image satisfies conditions based on the AI ​​inspection result 20R and the comprehensive general inspection result 10R. The image storage unit 30 allows the captured image P to be stored if the AIOK flag is assigned to the captured image P. This allows the system to selectively store effective images that meet the purpose, taking into account not only the captured image P that is judged to be pass in the comprehensive general inspection result 10R, but also the AI ​​inspection result 20R, thereby improving storage efficiency. Therefore, the visual inspection device 100 can store many effective inspection images.

[0038] The image storage method according to the embodiment includes a comprehensive general inspection step in which a captured image P of a workpiece is subjected to a general inspection based on multiple different types of inspection items, and an overall general inspection result 10R indicating whether the image is good or bad is output. An AI inspection step in which a trained machine learning model M is used to perform a pass / fail judgment based on the captured image P, and an AI inspection result 20R indicating whether the image is good or bad is output. An AIOK flag determination step S1 determines whether the captured image P has an AIOK flag if the captured image P satisfies preset image storage conditions. If the captured image P has an AIOK flag, the captured image P can be saved. This allows for selective saving of valid images that meet the purpose, taking into account not only the captured image P that is judged to be good in the comprehensive general inspection result 10R, but also the AI ​​inspection result 20R, thereby improving storage efficiency. This provides an image storage method that can save many valid inspection images.

[0039] The image storage program according to the embodiment includes a comprehensive general inspection function that causes a computer to perform a pass / fail judgment on a captured image P of a workpiece based on multiple general inspections of different types of inspection items and output a comprehensive general inspection result 10R indicating pass / fail; an AI inspection function that causes a trained machine learning model M to perform a pass / fail judgment based on the captured image P and output an AI inspection result 20R indicating pass / fail; a flag assignment function that assigns an AIOK flag to the captured image P if the captured image P meets preset image storage conditions; and an AIOK flag determination function that determines whether the captured image P has an AIOK flag if the captured image P meets the image storage conditions. If the AIOK flag determination function returns a YES result, the captured image P can be saved. This allows the computer to selectively save valid images that meet the purpose, taking into account not only the captured image P that is judged pass in the comprehensive general inspection result 10R, but also the AI ​​inspection result 20R, improving storage efficiency. This provides an image storage program that can save many valid inspection images. [Explanation of symbols]

[0040] 100 Visual inspection equipment 10. General Inspection Department 10R General Inspection Results 20 AI Inspection Department 20R AI test results 30 Image storage section 40 AI inspection condition determination unit M Machine Learning Model P Captured image S1 AIOK flag determination step S2 First comprehensive result determination step S3 Second comprehensive result judgment step S4 Image saving condition determination step

Claims

1. a comprehensive general inspection unit that performs a pass / fail judgment on a captured image of the workpiece by comprehensively carrying out general inspections using a plurality of different types of inspection items, and outputs a comprehensive general inspection result indicating pass or fail; An AI inspection unit that judges whether the image is good or bad based on a trained machine learning model that can be trained, and outputs an AI inspection result indicating good or bad; an image storage unit capable of storing the captured image, The AI ​​inspection unit assigns an AI OK flag to the captured image when the AI ​​inspection result has one or more good judgments, the AI ​​inspection result has no bad judgments, and the overall general inspection result has no bad judgments. The image storage unit enables selective storage of the captured image when an AIOK flag is assigned to the captured image. Visual inspection equipment.

2. The image storage unit stores the captured image when an image storage setting is ON. The visual inspection device according to claim 1 .

3. an inspection condition determination unit that outputs a determination result as to whether or not the average density of the inspection execution area in the captured image is within a preset range, and enables the AI ​​inspection unit to perform an operation when the determination result is within the set range; 3. The visual inspection apparatus according to claim 1 or 2.

4. a comprehensive general inspection step in which a pass / fail judgment is made on the captured image of the workpiece based on a plurality of different types of general inspection items, and a comprehensive general inspection result indicating pass or fail is output; An AI inspection step of determining whether the image is good or bad using a trained machine learning model that can be trained based on the captured image and outputting an AI inspection result indicating good or bad; an AIOK flag determination step of determining whether or not the captured image has an AIOK flag when the captured image satisfies a preset image storage condition, In the AI ​​inspection step, when the conditions that the AI ​​inspection result has one or more good judgments, the AI ​​inspection result has no bad judgments, and the overall general inspection result is not a bad judgment are satisfied, an AI OK flag is assigned to the captured image; If the captured image has the AIOK flag, the captured image can be selectively saved. How to save images.

5. If the determination in the AI ​​OK flag determination step is NO, a first overall result determination step is executed to determine whether the overall general inspection result is OK or not; If the comprehensive general examination result is OK, the captured image can be saved. The image storage method according to claim 4 .

6. If the first overall result determination step is NO, a second overall result determination step is executed to determine whether the overall general test result is NG or not; If the second overall result determination step determines YES, the captured image is made saveable. The image storage method according to claim 5 .

7. an image storage condition determination step for outputting a determination result as to whether or not the average density of the inspection execution area in the captured image is within a preset range; In the image storage condition determination step, if the determination result is within the set range, the AI ​​inspection step is executed.

7. The image storage method according to claim 4, wherein the image is stored in a storage medium.

8. On the computer, a comprehensive general inspection function that performs a pass / fail judgment on a captured image of a workpiece by comprehensively performing general inspections using a plurality of different types of inspection items, and outputs a comprehensive general inspection result indicating pass or fail; An AI inspection function that executes a pass / fail judgment based on the captured image using a trained machine learning model and outputs an AI inspection result indicating pass or fail; A flagging function that assigns an AI OK flag to the captured image when the captured image satisfies a preset image storage condition, the AI ​​inspection result has one or more good judgments, the AI ​​inspection result has no bad judgments, and the overall general inspection result has no bad judgments; and an AIOK flag determination function that determines whether the captured image has an AIOK flag when the captured image satisfies the image storage condition; When the AIOK flag determination function determines YES, the captured image can be selectively saved. Image saving program.

9. If the AI ​​OK flag determination function is NO, a first overall result determination function is executed to determine whether the overall general test result is OK or not; If the first overall result determination function determines YES, the captured image can be saved. The image saving program according to claim 8.

10. If the first overall result determination function is NO, a second overall result determination function is executed to determine whether the overall general test result is NG or not; If the second overall result determination function determines YES, the captured image can be saved. The image saving program according to claim 9 .

11. Execute an image storage condition determination function that outputs a determination result as to whether or not the average density of the inspection execution area in the captured image is within a preset range; In the image storage condition determination function, if the determination result is within the set range, the AI ​​inspection function is executed.

11. The image saving program according to claim 8.

Images