Information processing systems and programs

Abstract

To facilitate the performance of inspections on target images using thresholds set by the user as intended. [Solution] An information processing system comprising a processor, the processor acquires a source image, accepts user operations on the source image, generates a pseudo-abnormal image containing abnormal image features based on the acquired source image and user operations, sets the abnormality level of the pseudo-abnormal image obtained by inputting the pseudo-abnormal image to a learning model capable of calculating the abnormality level of an image when an image is input, sets the abnormality level of the pseudo-abnormal image as a threshold, acquires an image to be inspected, compares the abnormality level of the image to be inspected obtained by inputting the image to be inspected to the learning model with the set threshold, and performs an inspection on the image to be inspected.

Description

Technical Field

[0001] The present invention relates to an information processing system and a program.

Background Art

[0002] In Patent Document 1, based on the output of the intermediate layer of a learned neural network into which each of a plurality of reference images is input, a feature amount of the reference image corresponding to each of the plurality of reference images is extracted, and based on the output of the intermediate layer of the learned neural network into which an image of a subject to be inspected is input, a feature amount corresponding to the image is extracted, and based on these extracted feature amounts, an abnormality determination method for determining whether the appearance of the subject to be inspected is normal or not is described.

[0003] In Patent Document 2, identification information (M_thr1) for distinguishing normal image data (D_g) from at least one type of specific abnormal image data (D_ng) is set based on the output result when the normal image data (D_g) and the at least one type of specific abnormal image data (D_ng) are input into a learning model, and image data (D) of an inspection target product (W) is input into a normal product learning model, and an abnormality determination of the inspection target product (W) is performed based on the identification information (M_thr1). An abnormality determination method is described.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0005] Sometimes, an image is inspected by comparing its abnormality level with a threshold value. The abnormality level of the image is obtained by inputting the image into a learning model. However, setting the threshold value as intended by the user is difficult. In other words, it is not easy to perform an inspection of an image using a threshold value set as intended by the user.

[0006] The objective of this invention is to facilitate the performance of inspections on images to be inspected using thresholds set as intended by the user. [Means for solving the problem]

[0007] The invention described in claim 1 is an information processing system comprising a processor, the processor acquires a source image, accepts user operations on the source image, generates a pseudo-abnormal image containing abnormal image features based on the acquired source image and the user operations, sets the abnormality level of the pseudo-abnormal image obtained by inputting the pseudo-abnormal image to a learning model capable of calculating the abnormality level of an image when an image is input as a threshold, acquires an image to be inspected, compares the abnormality level of the image to be inspected obtained by inputting the image to be inspected to the learning model with the set threshold, and performs an inspection on the image to be inspected. The invention described in claim 2 is an information processing system according to claim 1, wherein the processor generates the pseudo-abnormal image containing different abnormal image features in response to different user operations. The invention described in claim 3 is an information processing system according to claim 2, wherein the inspection of the image to be inspected is an inspection of an object depicted in the image to be inspected, and the processor generates the pseudo-abnormal image containing different abnormal image features even with the same user operation, depending on the type of object depicted in the image to be inspected. The invention described in claim 4 is an information processing system according to claim 1, wherein the processor generates a plurality of pseudo-abnormal images, obtains the degree of abnormality of each of the plurality of pseudo-abnormal images, and sets the obtained degree of abnormality of each of the plurality of pseudo-abnormal images as a plurality of thresholds in the inspection. The invention described in claim 5 is an information processing system according to claim 4, wherein the processor generates a plurality of pseudo-abnormal images, each containing different degrees of abnormal image features for the same user operation, depending on which of the plurality of states the system is in. The invention described in claim 6 is an information processing system according to claim 1, wherein the inspection of the image to be inspected is an inspection of objects depicted in the image to be inspected, and the processor obtains the original image by performing processing according to the type of object depicted in the image to be inspected. The invention described in claim 7 is an information processing system according to claim 6, wherein the image to be inspected is an image obtained by scanning a printed document printed based on print data, and the processor acquires the print data as the original image. The invention described in claim 8 is an information processing system according to claim 1, wherein the processor adjusts the pseudo-abnormal image based on a first user operation on an image feature other than the abnormal one and a second user operation on the abnormal image feature. The invention described in claim 9 is an information processing system according to claim 8, wherein the processor is controlled to display a first display element indicating when it is possible to accept a first user operation, and is controlled to display a second display element indicating when it is possible to accept a second user operation. The invention described in claim 10 is a program for a computer to implement: a function to acquire an original image; a function to accept user operations on the original image; a function to generate a pseudo-abnormal image containing abnormal image features based on the acquired original image and the user operations; a function to set the degree of abnormality of the pseudo-abnormal image obtained by inputting the pseudo-abnormal image to a learning model capable of calculating the degree of abnormality of an image when an image is input, as a threshold; a function to acquire an image to be inspected; and a function to perform an inspection of the image to be inspected by comparing the degree of abnormality of the image to be inspected obtained by inputting the image to be inspected to the learning model with the set threshold. [Effects of the Invention]

[0008] According to the invention of claim 1, it becomes easier to perform inspections on the image to be inspected using thresholds set as intended by the user. According to the invention of claim 2, a threshold can be set using pseudo-abnormal images generated in various ways by various user operations. According to the invention of claim 3, the complexity of generating pseudo-abnormal images by performing different user operations for each type of object in the image to be inspected can be reduced. According to the invention of claim 4, it becomes easier to perform inspections on the image to be inspected using multiple thresholds set as intended by the user. According to the invention of claim 5, the complexity of generating pseudo-abnormal images by performing different user operations for each degree of abnormal image features can be reduced. According to the invention of claim 6, an appropriate image corresponding to the object in the image to be inspected can be obtained as the source image of the pseudo-abnormal image used to set a threshold. According to the invention of claim 7, when the image to be inspected is an image scanned from a printed document, the original image of the pseudo-abnormal image can be easily obtained. According to the invention of claim 8, the operability when adjusting the pseudo-abnormal image used to set a threshold can be improved. According to the invention of claim 9, it is possible to inform the user of what operations can be received when adjusting the pseudo-abnormal image used for setting the threshold value. According to the invention of claim 10, it becomes easy to perform an inspection on the inspection target image using the threshold value set as intended by the user.

Brief Description of the Drawings

[0009] [Figure 1] It is a diagram showing an example of the hardware configuration of the image inspection apparatus in the present embodiment. [Figure 2] It is a block diagram showing an example of the functional configuration of the image inspection apparatus in the present embodiment. [Figure 3] (a) and (b) are graphs showing the relationship between the abnormality degree of the inspection target image and the threshold value. [Figure 4] It is a diagram showing an example of a user operation for generating an abnormal image when the inspection on the inspection target image is a printed surface inspection. [Figure 5] It is a diagram showing a first example of a user operation for generating an abnormal image when the inspection on the inspection target image is a component inspection. [Figure 6] It is a diagram showing a second example of a user operation for generating an abnormal image when the inspection on the inspection target image is a component inspection. [Figure 7] (a) and (b) are diagrams showing a user operation when performing fine adjustment of an abnormal image. [Figure 8] It is a flowchart showing an example of the operation of the image inspection apparatus in the present embodiment.

Embodiments for Carrying Out the Invention

[0010] Hereinafter, the present embodiment will be described in detail with reference to the accompanying drawings.

[0011] (Outline of the Present Embodiment) This embodiment provides an information processing system that acquires an original image, receives a user operation on the original image, generates a pseudo-abnormal image based on the original image and the user operation, sets the abnormality degree of the pseudo-abnormal image as a threshold, acquires an inspection target image, and performs an inspection on the inspection target image by comparing the abnormality degree of the inspection target image with the threshold.

[0012] Here, the "original image" may be any image as long as it is the original of the pseudo-abnormal image. Hereinafter, a normal image will be taken as an example for explanation.

[0013] Also, the "system" may be composed of a single device or a plurality of devices. Hereinafter, as an information processing system, an example of a system composed of a single device will be taken. And, as the single device, an image inspection device will be taken as an example for explanation.

[0014] (Hardware Configuration of Image Inspection Device) FIG. 1 is a diagram showing an example of the hardware configuration of an image inspection device 10 in this embodiment. As shown in the figure, the image inspection device 10 includes a processor 11. The image inspection device 10 further includes a main memory 12 and an HDD (Hard Disk Drive) 13. The image inspection device 10 further includes a communication interface (I / F) 14, a display device 15, and an input device 16.

[0015] The processor 11 executes various software such as an OS (Operating System) and applications, and realizes each function described later. The main memory 12 is a memory used as a working memory of the processor 11 and the like. The HDD 13 stores input data for various software, output data from various software, and the like. The HDD 13 is, for example, a magnetic disk device. The communication I / F 14 transmits and receives various information to and from other devices via a communication line. The display device 15 displays various information. The display device 15 is, for example, a display. The input device 16 is used by the user to input information. The input device 16 is, for example, a keyboard or a mouse.

[0016] (Functional configuration of the imaging inspection device) Figure 2 is a block diagram showing an example of the functional configuration of the image inspection device 10 in this embodiment. As shown in the figure, the image inspection device 10 includes a normal image acquisition unit 21, a user operation reception unit 22, and an abnormal image generation unit 23. The image inspection device 10 further includes an abnormality degree acquisition unit 24, an inspection target image acquisition unit 25, and an inspection unit 26. These functional units are realized by the processor 11 reading a program from the HDD 13 into the main memory 12 and executing it.

[0017] The normal image acquisition unit 21 acquires a normal image and displays it on the display device 15. Here, a normal image is an image that has normal image characteristics in terms of image quality or content. A normal image may also be a group of normal images that includes multiple normal images, but in the following description, it will be described as a "normal image". In this embodiment, the processing of the normal image acquisition unit 21 is performed as an example of the process of acquiring the original image.

[0018] The user operation reception unit 22 receives user operations on the normal image displayed on the display device 15. User operations may be operations performed by the user on the display device 15 using the input device 16. In this embodiment, the processing of the user operation reception unit 22 is performed as an example of the process of receiving user operations on the original image.

[0019] The abnormal image generation unit 23 receives a normal image from the normal image acquisition unit 21 and user operations from the user operation reception unit 22. The abnormal image generation unit 23 then generates an abnormal image based on the normal image and the user operations. Here, an abnormal image is an image that has abnormal image features as image features related to image quality or content. Furthermore, this abnormal image is not an image that actually has abnormal image features, but an image that is generated to appear to have pseudo-abnormal image features. In this sense, this abnormal image is sometimes called a "pseudo-abnormal image". In this embodiment, the abnormal image generation unit 23 performs this process as an example of the process of generating a pseudo-abnormal image containing abnormal image features based on the acquired original image and user operations.

[0020] Furthermore, the abnormal image generation unit 23 may generate multiple abnormal images based on a normal image and user operations. In this embodiment, this process of the abnormal image generation unit 23 is performed as an example of the process of generating multiple pseudo-abnormal images. In this case, the image inspection device 10 may be set to a high abnormal grade mode and a low abnormal grade mode. In this case, suppose a certain user operation is performed. Then, if the high abnormal grade mode is set, the abnormal image generation unit 23 should add an abnormality of a high abnormal level. Also, if the low abnormal grade mode is set, the abnormal image generation unit 23 should add an abnormality of a low abnormal level. In this case, the high abnormal grade mode and the low abnormal grade mode are examples of multiple states. The processing of the abnormal image generation unit 23 is an example of the process of generating multiple pseudo-abnormal images that each contain different degrees of abnormal image features, even for the same user operation, depending on which of the multiple states the system is in.

[0021] The abnormality score acquisition unit 24 acquires the abnormality score of the abnormal image generated by the abnormal image generation unit 23. Specifically, the abnormality score acquisition unit 24 acquires the abnormality score obtained by inputting the abnormal image into the AI ​​(learning model). For example, the abnormality score acquisition unit 24 receives a normal image from the normal image acquisition unit 21 and an abnormal image from the abnormal image generation unit 23. The abnormality score acquisition unit 24 also acquires the feature vector of the normal image obtained by inputting the normal image into the AI ​​(learning model). The abnormality score acquisition unit 24 also acquires the feature vector of the abnormal image obtained by inputting the abnormal image into the AI ​​(learning model). Furthermore, the abnormality score acquisition unit 24 calculates the distance between these feature vectors and uses that distance as the abnormality score. The abnormality score acquisition unit 24 then sets this abnormality score as a threshold for determining whether the image to be inspected is normal or abnormal. Here, the AI ​​(learning model) may be located within the image inspection device 10. Alternatively, the AI ​​(learning model) may be located on a cloud server or the like outside the image inspection device 10. In this embodiment, the abnormality acquisition unit 24 performs this process as an example of setting the abnormality of a pseudo-abnormal image obtained by inputting a pseudo-abnormal image to a learning model that can calculate the abnormality of an image when an image is input, as a threshold.

[0022] Furthermore, when the abnormal image generation unit 23 generates multiple abnormal images, the abnormality acquisition unit 24 acquires multiple abnormality scores for these multiple abnormal images. Here, each of the multiple abnormality scores is the abnormality score for the corresponding abnormal image among the multiple abnormal images. In this embodiment, this process of the abnormality acquisition unit 24 is performed as an example of the process of acquiring the abnormality score for each of the multiple pseudo-abnormal images. The abnormality acquisition unit 24 then sets these multiple abnormality scores as thresholds for determining whether the image to be inspected is normal or to what extent it is abnormal. In this embodiment, this process of the abnormality acquisition unit 24 is performed as an example of the process of setting the acquired abnormality scores for each of the multiple pseudo-abnormal images as multiple thresholds in the inspection.

[0023] The inspection target image acquisition unit 25 acquires the inspection target image. Here, the inspection target image is the image to be inspected. The inspection target image may also be a group of inspection target images that includes multiple inspection target images, but in the following description, it will be described as "inspection target image". In this embodiment, the processing of the inspection target image acquisition unit 25 is performed as an example of the process of acquiring the inspection target image.

[0024] The inspection unit 26 obtains the abnormality score of the image to be inspected, which was acquired by the image acquisition unit 25. Specifically, the inspection unit 26 obtains the abnormality score obtained by inputting the image to be inspected into an AI (learning model). For example, the inspection unit 26 receives a normal image from the normal image acquisition unit 21 and an image to be inspected from the image to be inspected acquisition unit 25. The inspection unit 26 also obtains the feature vector of the normal image obtained by inputting the normal image into an AI (learning model). The inspection unit 26 obtains the feature vector of the image to be inspected obtained by inputting the image to be inspected into an AI (learning model). Furthermore, the inspection unit 26 calculates the distance between these feature vectors and uses that distance as the abnormality score of the image to be inspected. Then, the inspection unit 26 performs an inspection of the image to be inspected by comparing this abnormality score with a threshold set by the abnormality score acquisition unit 24. Here, the inspection of the image to be inspected refers to, for example, an inspection of the image quality or content of the image to be inspected. In this embodiment, the inspection unit 26 performs the processing shown in the example of an inspection process that compares the degree of abnormality of the image to be inspected, obtained by inputting the image to be inspected into a learning model, with a set threshold, to perform an inspection on the image to be inspected.

[0025] Figures 3(a) and 3(b) are graphs showing the relationship between the abnormality level of the images being examined and the threshold. In Figures 3(a) and 3(b), the abnormality levels of multiple images being examined are arranged in ascending order. Figure 3(a) shows a single threshold TH. This threshold TH is the threshold when the abnormal image generation unit 23 generates one abnormal image and the abnormality level acquisition unit 24 acquires one abnormality level. If the abnormality level of the image to be inspected is greater than or equal to threshold TH, the image to be inspected is determined to be abnormal. If the abnormality level of the image to be inspected is less than threshold TH, the image to be inspected is determined to be normal. Figure 3(b) shows multiple thresholds TH1 and TH2. These thresholds TH1 and TH2 are the thresholds used when the abnormal image generation unit 23 generates multiple abnormal images and the abnormality level acquisition unit 24 acquires multiple abnormality levels. If the abnormality level of the image under inspection is greater than or equal to threshold TH1, the image under inspection is determined to be at a high level of abnormality. If the abnormality level of the image under inspection is less than threshold TH1 but greater than or equal to threshold TH2, the image under inspection is determined to be at a low level of abnormality. If the abnormality level of the image under inspection is less than threshold TH2, the image under inspection is determined to be normal.

[0026] (Specific examples of user actions) Figure 4 shows an example of user operations that generate abnormal images when the inspection of the image to be inspected is a printed surface inspection. Initially, the normal image acquisition unit 21 is assumed to be displaying a normal image 300. Now, suppose the user performs a click (tap) operation as indicated by the finger mark 315. Then, the abnormal image generation unit 23 generates an abnormal image 310 with a dot 311. Alternatively, suppose the user performs an operation to draw a line while clicking, as indicated by the finger mark 325 and arrow 326. Then, the abnormal image generation unit 23 generates an abnormal image 320 with lines 321. Alternatively, suppose the user performs an operation where they click and move to fill in an area, as indicated by the finger mark 335 and the arrow 336. In this case, the abnormal image generation unit 23 generates an abnormal image 330 with unevenness 331.

[0027] Furthermore, the abnormal image generation unit 23 may control the size of the abnormality by operating the mouse wheel. Furthermore, the abnormal image generation unit 23 may control the intensity of the abnormality by dragging with a mouse.

[0028] Figure 5 shows a first example of user operations that generate an abnormal image when the inspection of the image to be inspected is a component inspection. Here, a metal nut is used as an example component. Initially, the normal image acquisition unit 21 is assumed to be displaying a normal image 400. Now, suppose the user performs an operation where they click and quickly move the cursor, as indicated by the finger mark 415 and arrow 416. Then, the abnormal image generation unit 23 generates an abnormal image 410 with a scratch defect 411. In this case, the abnormal image generation unit 23 may change the degree of the scratch defect 411 depending on the number of times the cursor is moved. Alternatively, suppose the user performs an operation of clicking and holding, as indicated by the finger mark 425. Then, the abnormal image generation unit 23 generates an abnormal image 420 with a dent defect 421. In this case, the abnormal image generation unit 23 may change the degree of indentation of the dent defect 421 depending on the duration of the long press or the strength of the tap.

[0029] Furthermore, the abnormal image generation unit 23 may control the size of the abnormality by operating the mouse wheel. Furthermore, the abnormal image generation unit 23 may control the intensity of the abnormality by dragging with a mouse.

[0030] Figure 6 shows a second example of user operations that generate abnormal images when the inspection of the image being inspected is a component inspection. In this example, a screw is used as the component. Initially, the normal image acquisition unit 21 is assumed to be displaying a normal image 500. Now, suppose the user selects region 550 and performs an operation to draw an illustration as shown by line 515. Then, the abnormal image generation unit 23 generates an abnormal image 510 having a missing tip portion 511 as shown in the illustration. Alternatively, assume that the user selects area 550 and performs an operation of inputting a voice or text saying "Broken" or "chipped". Then, the abnormal image generation unit 23 generates an abnormal image 520 having a chipped tip portion 521. However, simply saying "Broken" or "chipped" can result in various chipping patterns. Therefore, the abnormal image generation unit 23 may generate a plurality of abnormal images 520 having chipped tip portions 521 with various chipping patterns. Thereby, the user selects, from the plurality of abnormal images 520, an abnormal image 520 having a chipped tip portion 521 with a desired chipping pattern.

[0031] By the way, in FIGS. 4 to 6, the abnormal image generation unit 23 generates different abnormal images depending on different user operations. Here, different abnormal images are abnormal images having different abnormal image features. In that sense, the processing of the abnormal image generation unit 23 is an example of processing for generating a pseudo-abnormal image including different abnormal image features according to different user operations.

[0032] Also, in FIGS. 4 to 6, as inspections regarding the inspection target image, printing surface inspection or component inspection is assumed. Here, the printing surface and the component can be regarded as objects shown in the inspection target image. In that sense, the printing surface inspection and the component inspection are examples of inspections of objects shown in the inspection target image.

[0033] In that case, in FIGS. 4 to 6, the abnormal image generation unit 23 generates different abnormal images depending on the type of inspection even if the user operation is the same. That is, the abnormal image generation unit 23 generates different abnormal images depending on the type of object shown in the inspection target image even if the user operation is the same. For example, consider the case where the user performs an operation of drawing a line while clicking. In this case, if the inspection target is the printing surface as shown in FIG. 4, the abnormal image generation unit 23 generates an abnormal image 320 with a streak 321. On the other hand, if the inspection target is a component as shown in FIG. 5, the abnormal image generation unit 23 generates an abnormal image 420 with a scratch defect 411. In this case, the processing of the abnormal image generation unit 23 is an example of processing for generating a pseudo-abnormal image including different abnormal image features with the same user operation depending on the type of object shown in the inspection target image.

[0034] Furthermore, in Figures 4 to 6, the normal image acquisition unit 21 acquires a normal image by performing different processing depending on the type of inspection. In other words, the normal image acquisition unit 21 acquires a normal image by performing different processing depending on the type of object depicted in the image to be inspected. In this case, the processing of the normal image acquisition unit 21 is an example of the process of acquiring the original image by performing processing according to the type of object depicted in the image to be inspected. For example, consider the case where the object to be inspected is a printed surface, as shown in Figure 4. In this case, the normal image acquisition unit 21 should acquire the original print data of the printed surface as the normal image 300. Here, let's assume that the printed surface is an image obtained by scanning a printed document that was printed based on the print data. Then, the normal image acquisition unit 21 acquires that print data as the normal image 300. In this case, the processing of the normal image acquisition unit 21 is an example of the process of acquiring print data as the original image.

[0035] Figures 7(a) and 7(b) illustrate the user operations involved in fine-tuning an abnormal image. Here, abnormal image 420 shown in Figure 5 is used as an example, but these user operations can be applied to any abnormal image.

[0036] Figure 7(a) shows user operations on the background image of an abnormal image. Here, the background image is the image of the abnormal image other than the abnormal image features. In Figure 7(a), the user is performing an operation to enlarge the metal nut from the state of the abnormal image 420 in Figure 5. In this state, the user can also move the metal nut so that the dent defect 421 is positioned, for example, in the center of the image. Conversely, the user can also shrink the metal nut. The user may enlarge or reduce the background image by operating the mouse wheel, for example. The fact that these user operations on the background image are possible is indicated by a thick dashed frame 611 surrounding the image. The thick dashed frame 611 may be, for example, a red frame. In this case, user operations on the background image are an example of a first user operation on something other than the abnormal image features. The thick dashed frame 611 is an example of a first display element that indicates that the first user operation is possible.

[0037] Figure 7(b) shows user operations on abnormal image features of an abnormal image. In Figure 7(b), the user can change the size and color of the dent defect 421. The user may change the size of the dent defect 421 by, for example, operating the mouse wheel. Alternatively, the user may change the color of the dent defect 421 by, for example, writing characters on the display device 15 with the mouse. Here, the characters are exemplified by characters representing the colors and values ​​of "C", "M", "Y", and "K". Alternatively, the user may change the color of the dent defect 421 by, for example, inputting each color and its value using the keyboard. The fact that these user operations on abnormal image features are possible is indicated by a thick solid line frame 621 surrounding the image. The thick solid line frame 621 may be, for example, a green frame. In this case, the user operations on the abnormal image features are an example of a second user operation on abnormal image features. The thick solid line frame 621 is an example of a second display element indicating that a second user operation is possible.

[0038] Furthermore, if the area surrounding the abnormal image feature is clicked (tapped) in the state shown in Figure 7(b), the image may transition to the state shown in Figure 7(a).

[0039] When the user performs these operations, the abnormal image generation unit 23 fine-tunes the abnormal image based on these operations. At that time, the abnormal image generation unit 23 displays a thick dashed frame 611 or a thick solid frame 621 indicating which of these operations is acceptable. In this case, the processing of the abnormal image generation unit 23 is an example of processing that adjusts the pseudo-abnormal image based on the first user operation and the second user operation. The processing of the abnormal image generation unit 23 is an example of processing that controls the display of the first display element and controls the display of the second display element.

[0040] (Operation of the imaging inspection device) Figure 8 is a flowchart showing an example of the operation of the image inspection device 10 in this embodiment.

[0041] As shown in the figure, in the image inspection device 10, first, the normal image acquisition unit 21 acquires a normal image (step 201). As a result, the normal image is displayed on the display device 15. Here, a normal image is an image that has normal image characteristics in terms of image quality or content.

[0042] Next, in the image inspection device 10, the user operation reception unit 22 receives user operations on a normal image (step 202). User operations may be operations performed by the user on the display device 15 using the input device 16.

[0043] Next, in the image inspection device 10, the abnormal image generation unit 23 generates an abnormal image (step 203). Specifically, the abnormal image generation unit 23 holds the normal image acquired in step 201 and the user operation received in step 202. Then, the abnormal image generation unit 23 generates an abnormal image based on the normal image and the user operation. Here, an abnormal image is an image that has abnormal image features as image features related to image quality or content.

[0044] Next, in the image inspection device 10, the abnormality acquisition unit 24 acquires the abnormality of the abnormal image generated in step 203 (step 204). Specifically, the abnormality acquisition unit 24 acquires the abnormality by inputting the abnormal image into the AI ​​(learning model). The abnormality acquisition unit 24 then sets this abnormality level as a threshold for determining whether the image to be inspected is normal or abnormal (step 205).

[0045] Meanwhile, in the image inspection device 10, the inspection target image acquisition unit 25 acquires the inspection target image (step 206). Here, the inspection target image is the image of the object to be inspected.

[0046] Next, in the image inspection device 10, the inspection unit 26 obtains the abnormality score of the image to be inspected acquired in step 206 (step 207). Specifically, the inspection unit 26 obtains the abnormality score obtained by inputting the image to be inspected into the AI ​​(learning model). Then, the inspection unit 26 performs an inspection on the image to be inspected (step 208). Specifically, the inspection unit 26 performs the inspection by comparing this abnormality level with the threshold set in step 205. Here, the inspection on the image to be inspected refers to, for example, an inspection of the image quality or content of the image to be inspected.

[0047] (Processor) In this embodiment, each process is executed on any computer. Furthermore, any computer may execute these processes using a processor as hardware, a program as software, or a combination thereof. In that case, the processor is configured to work in cooperation with the program to execute the various processes in this embodiment, and can function as a unit or means in this embodiment. Also, the execution order of the processes by the processor is not limited to the order described and may be changed as appropriate. Any computer may be a general-purpose computer, a computer designed for a specific purpose, a workstation, or any other system capable of executing each process.

[0048] A processor may consist of one or more hardware components, and the type of hardware is not limited. For example, a processor may consist of a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a programmable logic device such as an FPGA (Field Programmable Gate Array), a dedicated circuit for executing a specific process such as an ASIC (Application Specific Integrated Circuit), a GPU (Graphic Processing Unit), or an NPU (Neural Processing Unit). Furthermore, the type of hardware may be a combination of different types of hardware. When multiple hardware components are configured to execute one or more processes of a given processor, these components may reside in physically separate devices or in the same device. Also, in any embodiment, the order of each process performed by the processor is not limited to the order described above and may be changed as appropriate. Hardware is composed of electrical circuits (circuitry) that combine circuit elements such as semiconductor elements.

[0049] Furthermore, the program may be firmware or software such as microcode. Alternatively, the program may be, for example, a set of program modules, each function of which may be implemented by a processor configured to perform its respective function. The program may be program code or multiple code segments stored on one or more non-temporary computer-readable media (e.g., storage media or other storage). The program may be divided and stored on multiple non-temporary computer-readable media located on physically separate devices. Program code or code segments may represent any combination of procedures, functions, subprograms, routines, subroutines, modules, software packages, classes, or instructions, data structures, or program statements. Program code or code segments may be connected to other code segments or hardware circuits by sending and receiving information, data, arguments, parameters, or memory contents.

[0050] (program) This invention can also be applied to programs and program products. For example, a program and program product to which this embodiment is applied provides a computer with the following functions: a function to acquire an original image; a function to accept user operations on the original image; a function to generate a pseudo-abnormal image containing abnormal image features based on the acquired original image and user operations; a function to set the abnormality level of the pseudo-abnormal image obtained by inputting the pseudo-abnormal image to a learning model capable of calculating the abnormality level of an image when an image is input, as a threshold; a function to acquire an image to be inspected; and a function to perform an inspection of the image to be inspected by comparing the abnormality level of the image to be inspected obtained by inputting the image to be inspected to the learning model with the set threshold. Furthermore, the program to which this embodiment applies can be provided via communication means. Alternatively, the program to which this embodiment applies can be provided on a recording medium such as a CD-ROM.

[0051] (Note) (((1))) Equipped with a processor, The aforementioned processor, Get the original image, The system accepts user operations on the aforementioned original image. Based on the acquired original image and the user operations, a pseudo-abnormal image containing abnormal image features is generated. The abnormality level of the pseudo-abnormal image obtained by inputting the pseudo-abnormal image to a learning model capable of calculating the abnormality level of an image when an image is input is set as the threshold. Acquire the image to be examined, An information processing system that performs an inspection on an image by comparing the degree of abnormality of the image to be inspected, obtained by inputting the image to be inspected into the learning model, with the set threshold. (((2))) The information processing system according to (((1))), wherein the processor generates the pseudo-abnormal image containing different abnormal image features in response to different user operations. (((3))) The inspection of the aforementioned image subject to inspection is an inspection of the objects depicted in the image subject to inspection. The information processing system according to (((2))), wherein the processor generates the pseudo-abnormal image containing different abnormal image features even with the same user operation, depending on the type of object captured in the image to be inspected. (((4))) The processor generates multiple pseudo-abnormal images, The degree of abnormality of each of the aforementioned multiple pseudo-abnormal images is obtained, The degree of abnormality of each of the acquired multiple pseudo-abnormal images is set as multiple thresholds in the inspection. An information processing system as described in any of (((1))) to (((3))). (((5))) The information processing system according to (((4))), wherein the processor generates a plurality of pseudo-abnormal images, each containing different degrees of abnormal image features for the same user operation, depending on which of the multiple states the system is in. (((6))) The inspection of the aforementioned image subject to inspection is an inspection of the objects depicted in the image subject to inspection. The information processing system according to any one of (((1))) to (((5))), wherein the processor acquires the original image by performing processing according to the type of object captured in the image to be inspected. (((7))) The image subject to inspection is an image obtained by scanning a printed document that was printed based on print data. The information processing system described in (((6))) wherein the processor acquires the print data as the original image. (((8))) The information processing system according to any one of (((1))) to (((7))), wherein the processor adjusts the pseudo-abnormal image based on a first user operation on an image feature other than the abnormal one and a second user operation on the abnormal image feature. (((9))) The information processing system according to (((8))), wherein the processor is controlled to display a first display element indicating when it is possible to accept the first user operation, and is controlled to display a second display element indicating when it is possible to accept the second user operation. (((10))) On the computer, A function to retrieve the original image, A function that accepts user operations on the aforementioned original image, A function to generate a pseudo-abnormal image containing abnormal image features based on the acquired original image and the user operation, A function that sets the abnormality level of a pseudo-abnormal image obtained by inputting the pseudo-abnormal image to a learning model capable of calculating the abnormality level of an image when an image is input, as a threshold, A function to acquire images to be examined, A function that performs an inspection on the target image by comparing the degree of abnormality of the target image obtained by inputting the target image to the learning model with the set threshold, A program to achieve this.

[0052] According to the invention of (((1))), it becomes easy to perform inspections on the image to be inspected using thresholds set as intended by the user. According to the invention of (((2))), a threshold can be set using pseudo-abnormal images generated in various ways by various user operations. According to the invention of (((3))), the complexity of generating pseudo-abnormal images by performing different user operations for each type of object in the image to be inspected can be reduced. According to the invention of (((4))), it becomes easier to perform inspections on the image to be inspected using multiple thresholds set as intended by the user. According to the invention of (((5))), the complexity of generating pseudo-abnormal images by performing different user operations for each degree of abnormal image features can be reduced. According to the invention of (((6))), an appropriate image corresponding to the object in the image to be inspected can be obtained as the source image for the pseudo-abnormal image used to set a threshold. According to the invention of (((7))), when the image to be inspected is an image scanned from a printed document, the original image of the pseudo-abnormal image can be easily obtained. According to the invention of (((8))), the operability when adjusting the pseudo-abnormal image used to set a threshold can be improved. According to the invention of (((9))), the user can be informed of what operations are possible when adjusting the pseudo-abnormal image used to set a threshold. According to the invention of (((10))), it becomes easy to perform inspections on the image to be inspected using thresholds set as intended by the user. [Explanation of Symbols]

[0053] 10...Image inspection device, 21...Normal image acquisition unit, 22...User operation reception unit, 23...Abnormal image generation unit, 24...Abnormality level acquisition unit, 25...Inspection target image acquisition unit, 26...Inspection unit

Claims

1. Equipped with a processor, The aforementioned processor, Get the original image, The system accepts user operations on the aforementioned original image. Based on the acquired original image and the user operation, a pseudo-abnormal image containing abnormal image features is generated. The abnormality level of the pseudo-abnormal image obtained by inputting the pseudo-abnormal image to a learning model capable of calculating the abnormality level of an image when an image is input is set as the threshold. Acquire the image to be examined, An information processing system that performs an inspection on an image by comparing the degree of abnormality of the image to be inspected, obtained by inputting the image to be inspected into the learning model, with the set threshold.

2. The information processing system according to claim 1, wherein the processor generates the pseudo-abnormal image containing different abnormal image features in response to different user operations.

3. The inspection of the aforementioned image subject to inspection is an inspection of the objects depicted in the image subject to inspection. The information processing system according to claim 2, wherein the processor generates the pseudo-abnormal image containing different abnormal image features even with the same user operation, depending on the type of object captured in the image to be inspected.

4. The aforementioned processor generates multiple pseudo-anomalous images, The degree of abnormality of each of the aforementioned multiple pseudo-abnormal images is obtained, The degree of abnormality of each of the acquired multiple pseudo-abnormal images is set as one of several thresholds in the inspection. The information processing system according to claim 1.

5. The information processing system according to claim 4, wherein the processor generates a plurality of pseudo-abnormal images, each containing different degrees of abnormal image features for the same user operation, depending on which of the plurality of states the system is in.

6. The inspection of the aforementioned image subject to inspection is an inspection of the objects depicted in the image subject to inspection. The information processing system according to claim 1, wherein the processor acquires the original image by performing processing according to the type of object captured in the image to be inspected.

7. The image subject to inspection is an image obtained by scanning a printed document that was printed based on print data. The information processing system according to claim 6, wherein the processor acquires the print data as the original image.

8. The information processing system according to claim 1, wherein the processor adjusts the pseudo-abnormal image based on a first user operation on an image feature other than the abnormal one and a second user operation on the abnormal image feature.

9. The information processing system according to claim 8, wherein the processor is controlled to display a first display element indicating when it is possible to accept the first user operation, and is controlled to display a second display element indicating when it is possible to accept the second user operation.

10. On the computer, A function to retrieve the original image, A function that accepts user operations on the aforementioned original image, A function to generate a pseudo-abnormal image containing abnormal image features based on the acquired original image and the user operation, A function that sets the abnormality level of a pseudo-abnormal image obtained by inputting the pseudo-abnormal image to a learning model capable of calculating the abnormality level of an image when an image is input, as a threshold, A function to acquire images to be examined, A function that performs an inspection on the target image by comparing the degree of abnormality of the target image obtained by inputting the target image to the learning model with the set threshold, A program to achieve this.

Images