Information processing device, information processing method, and program

The system addresses the challenge of data-intensive anomaly detection by using statistical processing on normal images to generate detection criteria, achieving high-accuracy anomaly detection with reduced data requirements.

JP2024157084AInactive Publication Date: 2024-11-07BLENDING TECH CO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2023071192
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-04-25
Publication Date
2024-11-07
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing machine learning technologies for anomaly detection in manufacturing require a large amount of data, including abnormal data, and struggle with accuracy when insufficient data is available.

Method used

An information processing system that utilizes statistical processing on normal images to calculate an abnormality score and generate detection criteria, enabling high-accuracy anomaly detection without requiring abnormal images.

Benefits of technology

The system achieves high-accuracy anomaly detection using a smaller amount of data, reducing the need for extensive data collection and lowering costs, while maintaining or improving detection performance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024157084000001_ABST
    Figure 2024157084000001_ABST
Patent Text Reader

Abstract

To provide a technique capable of detecting abnormalities with high accuracy from a small amount of data.SOLUTION: A determination object image acquisition unit 60 acquires determination object image information transmitted from a factory terminal 3. A determination criterion acquisition unit 61 acquires determination criterion information related to a threshold value that is a determination criterion for determining whether there is an abnormality in a determination object image transmitted from a learning server 2. A determination unit 62 determines whether there is an abnormality in the determination object image based on an anomaly score calculated from the determination object image information acquired by the determination object image acquisition unit 60 and the determination criterion information acquired by the determination criterion acquisition unit 61. A determination result management unit 63 transmits, to the factory terminal 3, determination result information related to results of the determination made by the determination unit 62 on the determination object image.SELECTED DRAWING: Figure 3
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an information processing device, an information processing method, and a program. [Background technology]

[0002] In recent years, there have been remarkable advances in machine learning technology, and attempts are being made to use machine learning technology in manufacturing and other fields. In this regard, for example, a technology called an autoencoder has been proposed for detecting abnormal products using the results of learning performed using only images of normal products (see Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2022-085429 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, the technology described in the above-mentioned Patent Document 1 alone requires a very large amount of data for learning, and even if sufficient data can be collected, if there is no or only a small amount of abnormal data, it is not possible to determine an appropriate standard, and the accuracy of the judgment is not very high.

[0005] The present invention has been made in consideration of the above circumstances, and has an object to provide a technology capable of detecting anomalies with high accuracy through simpler learning.

[0006] In order to achieve the above object, an information processing device according to one aspect of the present invention comprises: a determination target image acquisition means for acquiring an image including a determination target as a determination target image; A criterion acquisition means for acquiring information regarding a criterion determined based on a score calculated by applying a statistical process to a learning image composed of normal images; an anomaly detection means for determining whether the image to be judged is a normal image or an abnormal image based on the judgment criterion; Equipped with.

[0007] An information processing method and a program according to one aspect of the present invention are also provided as an information processing method or a program corresponding to the information processing device according to one aspect of the present invention. Effect of the Invention

[0008] According to the present invention, it is possible to provide a technology that can detect anomalies with high accuracy through simpler learning. [Brief description of the drawings]

[0009] [Figure 1] 1 is a diagram illustrating an example of a configuration of an information processing system according to an embodiment of the present invention. [Diagram 2] 2 is a diagram illustrating an example of a hardware configuration of an inference server that constitutes an information processing system according to an embodiment of the present invention. FIG. [Diagram 3] 1 is a diagram illustrating an example of the functional configuration of an inference server, a learning server, a factory terminal, and an imaging device that constitute an information processing system according to an embodiment of the present invention. [Figure 4] FIG. 2 is a diagram illustrating an example of a method in which a learning server constituting an information processing system according to an embodiment of the present invention generates a judgment criterion (threshold value). [Diagram 5] 4 is a diagram illustrating an example of the flow of a criterion generation process among the processes executed by the learning server having the functional configuration of FIG. 3. [Figure 6] 4 is a diagram illustrating an example of the flow of a determination process among the processes executed by an inference server having the functional configuration of FIG. 3. [Figure 7]FIG. 5 is a diagram for explaining an example of a method in which a learning server constituting an information processing system according to an embodiment of the present invention generates a judgment criterion (threshold value), and is a diagram showing an example different from the example of FIG. [Figure 8] FIG. 8 is a diagram for explaining an example of a method in which a learning server constituting an information processing system according to one embodiment of the present invention generates a judgment criterion (threshold value), and is a diagram showing an example different from the examples of FIGS. 4 and 7. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0011] <Summary> First, FIG. 1 is a diagram showing an example of the configuration of an information processing system according to an embodiment of the present invention. 1 includes an inference server 1, a learning server 2, a factory terminal 3, and an imaging device 4. The inference server 1, the learning server 2, and the factory terminal 3 are connected to each other via a predetermined network N such as the Internet. Note that the network N is not a required component, and for example, NFC (Near Field Communication), Bluetooth (registered trademark), LAN (Local Area Network), etc. may be used.

[0012] The inference server 1 is configured with a general-purpose PC (Personal Computer) or the like, and is managed by an administrator of this system, etc. The inference server 1 applies a judgment criterion (threshold value) to an image (hereinafter referred to as a "judgement target image") that captures a situation such as a factory that is a target of anomaly detection, and executes various processes to judge whether or not an abnormality exists.

[0013] The learning server 2 is configured with a general-purpose PC (Personal Computer) or the like, and is managed by an administrator of this system, etc. The learning server 2 applies various processes, including statistical machine learning, to normal image data for learning acquired in advance (hereinafter referred to as "learning images"), and executes various processes for determining judgment criteria (thresholds) described later.

[0014] The factory terminal 3 is configured with a general-purpose PC (Personal Computer), tablet, smartphone, etc., and is managed by a user who receives the service related to this system (hereinafter referred to as the "anomaly detection service"). The factory terminal 3 acquires various image data including learning images and judgment target images, etc., and transmits each information to the inference server 1 or the learning server 2. The imaging device 4 is composed of various video cameras and the like, and for example, one or a plurality of imaging devices 4 are installed at various locations in the factory where the factory terminal 3 is installed.

[0015] <Hardware configuration> FIG. 2 is a diagram showing an example of a hardware configuration that constitutes an inference server of an information processing system according to an embodiment of the present invention. As shown in FIG. 2, the inference server 1 includes a control unit 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a bus 14, an input / output interface 15, an output unit 16, an input unit 17, a memory unit 18, a communication unit 19, and a drive 20.

[0016] The control unit 11 is composed of a microcomputer including a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), and a semiconductor memory, and executes various processes according to a program recorded in the ROM 22 or a program loaded from the memory unit 18 to the RAM 13. The RAM 13 also stores information necessary for the control unit 11 to execute various processes as appropriate.

[0017] The control unit 11, ROM 12, and RAM 13 are connected to one another via a bus 14. An input / output interface 15 is also connected to this bus 14. An output unit 16, an input unit 17, a storage unit 18, a communication unit 19, and a drive 20 are connected to the input / output interface 15.

[0018] The output unit 16 is composed of various liquid crystal displays, speakers, etc., and outputs various information as images and sounds.

[0019] The input unit 17 is composed of a keyboard, a mouse, etc., and is used to input various information.

[0020] The storage unit 18 is configured with a hard disk drive (HDD), a solid state drive (SSD), etc., and stores various data. In this embodiment, for example, various information including various programs and various databases is stored.

[0021] The communication unit 19 controls communications with other devices via a network N including the Internet.

[0022] The drive 20 is provided as necessary. Removable media 31, which may be a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is appropriately mounted in the drive 20. The program read from the removable media 31 by the drive 20 is installed in the storage unit 18 as necessary. The removable media 31 can also store various data stored in the storage unit 18 in the same manner as the storage unit 18.

[0023] The hardware configurations of the learning server 2 and the factory terminal 3 can be basically the same as the hardware configuration of the inference server 1 described above, so a description thereof will be omitted.

[0024] FIG. 3 is a diagram showing an example of the functional configuration of an inference server, a learning server, a factory terminal, and an imaging device that constitute an information processing system according to an embodiment of the present invention.

[0025] As shown in FIG. 3, in the control unit 11 of the inference server 1, a judgment target image acquisition unit 60, a judgment criterion acquisition unit 61, a judgment unit 62, and a judgment result management unit 63 function by executing various programs and the like.

[0026] The judgment target image acquisition section 60 of the inference server 1 acquires an image including a judgment target as a judgment target image. Specifically, the judgment target image acquisition unit 60 acquires, via the communication unit 19, information on an image that is to be judged for anomaly detection and that is transmitted from the factory terminal 3 (hereinafter, referred to as “judgment target image information”). The judgment target image acquisition unit 60 of the inference server 1 calculates a score (hereinafter referred to as the "anomaly score") as to whether the target image includes an anomaly or not based on the acquired judgment target image information. The details of the method of calculating the anomaly score will be described later together with the explanation of the functional configuration of the learning server 2.

[0027] The judgment criterion acquisition unit 61 of the inference server 1 acquires information about the judgment criterion determined based on the score calculated by applying statistical processing to the learning images constituted by normal images. Specifically, the judgment criteria acquisition unit 61 acquires information regarding the threshold value that serves as the judgment criterion for determining whether or not there is an abnormality in the judgment target image transmitted from the learning server 2 via the communication unit 19 (hereinafter referred to as ``judgment criteria information'').

[0028] The judgment unit 62 of the inference server 1 judges whether the judgment target image is a normal image or an abnormal image based on the judgment criterion. Specifically, the judgment unit 62 judges whether the judgment target image has an abnormality or not based on the abnormality score calculated from the judgment target image information acquired by the judgment target image acquisition unit 60 and the judgment criterion information acquired by the judgment criterion acquisition unit 61. The abnormality score will be described in detail later.

[0029] The judgment result management section 63 of the inference server 1 transmits information on the result of judgment made by the judgment section 62 on the judgment target image (hereinafter referred to as “judgment result information”) to the factory terminal 3.

[0030] As shown in FIG. 3, in the control unit 200 of the learning server 2, a learning image acquisition unit 220, a learning unit 221, and a criterion generation unit 222 function by executing various programs and the like. The learning image acquisition unit 220 of the learning server 2 acquires normal images as learning images. Specifically, the learning image acquisition unit 220 acquires information about the learning images transmitted from the factory terminal 3 (hereinafter, referred to as “learning image information”) via the communication unit 210. Here, one of the objects of the present invention is to perform highly accurate anomaly detection by learning from as few images of normal times as possible without using images of abnormal times. That is, the learning images referred to here are, for example, composed of a relatively small number of normal images that do not include images of abnormal times. Usually, in manufacturing sites and the like, it is difficult to obtain image data of abnormal times, which often becomes a hindrance when applying techniques such as machine learning. In order to solve such problems, the present invention has considered a method for detecting anomalies in factories and the like with high accuracy from learning data composed only of normal images and a relatively small amount of data.

[0031] The learning unit 221 of the learning server 2 applies statistical processing to the learning images that do not contain any abnormal images and are made up of only normal images, and calculates a score. Specifically, the learning unit 221 applies a predetermined statistical process to the learning image information acquired by the learning image acquisition unit 220, and calculates an anomaly score for the target image. Here, the statistical processing applied by the learning unit 221 may include, for example, a general-purpose algorithm for anomaly detection, such as AutoEncoder, PatchCore, PaDiM, etc. The learning unit 221 can apply or utilize any such algorithm to calculate an anomaly score from a predetermined image. For example, the AutoEncoder can be referred to in the above-mentioned Patent Document 1. Similarly, for example, PatchCore can be referred to in the below-mentioned Non-Patent Document 1. [Non-Patent Document 1] “Towards Total Recall in Industrial Anomaly Detection”,Karsten Roth, Latha Pemula, Joaquin Zepeda, Bernhard Scholkopf, Thomas Brox, Peter Gehler (CVPR2022)URL:https: / / openaccess.thecvf.com / content / CVPR2022 / papers / Roth_Towards_Total_Recall_in_Industrial_Anomaly_Detection_CVPR_2022_paper.pdf

[0032] The criterion generating unit 222 of the learning server 2 generates a criterion for anomaly detection based on the calculated score. Specifically, the judgment criterion generating unit 222 generates a judgment criterion (threshold) for making a judgment on the judgment target image by applying a predetermined statistical process to the anomaly scores of the learning images calculated by the learning unit 221. Note that an example of a specific method in which the judgment criterion generating unit 222 generates the judgment criterion will be described later with reference to FIG. 4 etc. Moreover, the criterion generating unit 222 transmits criterion information relating to the generated criterion to the inference server 1.

[0033] As shown in FIG. 3, in the control unit 300 of the factory terminal 3, an image management unit 320 and a judgment result acquisition unit 321 function by executing various programs and the like. The image management unit 320 of the factory terminal 3 acquires and manages information relating to various images captured by the imaging device 4. Specifically, in the pre-learning stage, the image management unit 320 acquires learning image information on the learning image captured by the imaging device 4, and transmits the information to the learning server 2. Then, in the inference stage after learning, the image management unit 320 acquires judgment target image information on the judgment target image captured by the imaging device 4, and transmits the information to the inference server 1.

[0034] The determination result acquisition unit 321 of the factory terminal 3 acquires the determination result information transmitted from the inference server 1 via the communication unit 310 .

[0035] Next, an example of a specific process in which the criterion generating unit 222 generates a criterion will be described with reference to Fig. 4. Fig. 4 is a diagram for explaining an example of a method in which a learning server constituting an information processing system according to an embodiment of the present invention generates a criterion.

[0036] Here, the learning unit 221 of the learning server 2 can calculate anomaly scores only from learning images of the factory under normal conditions by applying algorithms such as AutoEncoder, PatchCore, and PaDiM, as described above. In response to this, the criterion generating unit 222 considers the frequency distribution of the anomaly scores calculated from each of the learning images, For example, the criterion generating unit

number

[0037] First, an example of the flow of the criterion process executed by the learning server 2 will be described with reference to Fig. 5. Fig. 5 is a diagram illustrating an example of the flow of the criterion generation process among the processes executed by the learning server having the functional configuration of Fig. 3.

[0038] In step S1, the learning image acquisition unit 220 of the learning server 2 acquires, via the communication unit 210, learning image information related to learning images transmitted from the factory terminal 3.

[0039] In step S2, the learning unit 221 of the learning server 2 applies a predetermined statistical processing to the learning image information acquired by the learning image acquisition unit 220, and calculates an anomaly score regarding whether or not the target image contains an anomaly.

[0040] In step S3, the judgment criterion generation unit 222 of the learning server 2 generates a judgment criterion (threshold value) for making a judgment on the judgment target image by applying a predetermined statistical processing to the anomaly score of the learning image calculated by the learning unit 221.

[0041] In step S4, the criterion generating unit 222 of the learning server 2 transmits criterion information related to the generated criterion to the inference server 1. This ends the criterion generation process of the learning server 2.

[0042] Next, an example of the flow of the determination process executed by the inference server 1 will be described with reference to Fig. 6. Fig. 6 is a diagram for explaining an example of the flow of the determination process among the processes executed by the inference server having the functional configuration of Fig. 3.

[0043] In step S21, the judgment target image acquisition unit 60 of the inference server 1 acquires, via the communication unit 19, judgment target image information relating to the judgment target image transmitted from the factory terminal 3.

[0044] In step S22, the judgment target image acquisition section 60 of the inference server 1 calculates the abnormality score of the judgment target image based on the acquired judgment target image information.

[0045] In step S23, the judgment criterion acquisition unit 61 of the inference server 1 acquires, via the communication unit 19, judgment criterion information regarding a threshold value, which is the judgment criterion for determining whether or not there is an abnormality in the image to be judged transmitted from the learning server 2.

[0046] In step S24, the judgment unit 62 of the inference server 1 judges whether or not there is an abnormality in the image to be judged based on the abnormality score calculated from the image to be judged information acquired by the image to be judged acquisition unit 60 and the judgment criterion information acquired by the judgment criterion acquisition unit 61.

[0047] In step S25, the judgment result management section 63 of the inference server 1 transmits judgment result information on the result of judgment made by the judgment section 62 on the judgment target image to the factory terminal 3. This ends the judgment process of the inference server 1.

[0048] Although one embodiment of the present invention has been described above, the present invention is not limited to the above-described embodiment, and modifications, improvements, etc. that can achieve the object of the present invention are included in the present invention.

[0049] Here, another method in which the learning server 2 generates a judgment criterion (threshold) will be described with reference to Figures 7 and 8. Figures 7 and 8 are both diagrams showing another example of a method in which the learning server constituting the information processing system according to one embodiment of the present invention generates a judgment criterion (threshold).

[0050] In the above-mentioned embodiment (particularly the embodiment in FIG. 4), an example was described in which a judgment criterion (threshold) was generated based on the distribution of abnormality scores calculated only from learning images showing normal situations. In contrast, in the embodiment shown in FIG. 7 and FIG. 8, learning images showing pseudo-abnormal situations (hereinafter referred to as "pseudo-abnormal images") are created from learning images showing normal situations, an abnormality score is calculated from learning image information showing normal situations and information on the pseudo-abnormal images (hereinafter referred to as "pseudo-abnormal image information"), and a judgment criterion (threshold) is generated from the calculated abnormality score.

[0051] Specifically, as shown in FIG. 7, for example, a part of an image showing a normal situation can be tampered with by a method such as copy and paste to generate a learning image that is similar to a pseudo-abnormal situation. Then, as shown in FIG. 8, a judgment criterion (threshold) can be generated from a distribution of anomaly scores calculated from images including normal learning images and pseudo-abnormal images. The learning server 2 can also generate a judgment criterion (threshold) by using such a method. Note that in the embodiments of FIG. 7 and FIG. 8, the applicants of the present invention have changed various combinations and conditions to examine the judgment criterion (threshold) so as to increase the accuracy of detecting anomalies (high F1 score, no oversight of abnormal images, etc.).

[0052] Although the above embodiment has been described simply, the usefulness of this system will be described again. (1) This system can detect anomalies only from normal images, without including abnormal images in the learning images. (2) Compared to existing algorithms, including those mentioned above, this system is expected to be able to detect anomalies using a smaller amount of data (learning image information). (3) This system is expected to be able to detect anomalies with high accuracy (despite using a small amount of data) compared to existing algorithms, including those mentioned above. (4) In conventional methods for detecting anomalies, a method for setting a threshold value and detecting anomalies based only on images of normal conditions is not believed to be disclosed, including the above-mentioned algorithm. (5) Even after generating the judgment criteria (thresholds) once, the accuracy of the system can be improved by adding pseudo-abnormal images (the embodiments shown in Figures 7 and 8) and regenerating the judgment criteria (thresholds). Due to these effects, this system can reduce the cost of collecting learning data and make it easy to build an anomaly detection system, making it expected to be introduced to the visual inspection of a variety of products.

[0053] Furthermore, the criteria for generating the judgment criteria (thresholds) described in the above-mentioned embodiment (particularly the embodiment in FIG. 4) are merely examples and are not limiting. The present system does not necessarily need to set the average value + 3σ as the threshold, and may set the threshold value based on any criteria.

[0054] Although not described in the above embodiment, the present system may have a function of changing the judgment criterion (threshold) in consideration of the result of judgment made by the human when the result of judgment made by the present system differs from the result of judgment made by the human, thereby enabling the present system to detect abnormal images from the judgment target images with even greater accuracy.

[0055] Although not described in the above embodiment, the judgment by the inference server 1 does not have to be performed only once, but may be performed periodically multiple times at regular intervals or in response to a predetermined trigger. For example, the factory terminal 3 may constantly transmit judgment target image information to the inference server 1 every five minutes and obtain the judgment result.

[0056] In the above embodiment, the present system has been described as employing a general-purpose algorithm for anomaly detection, such as AutoEncoder, PatchCore, PaDiM, etc., and calculating an anomaly score from normal data, but is not limited thereto. The present system may employ any algorithm other than these algorithms or any algorithm that is an improvement of these algorithms, and use it to calculate an anomaly score.

[0057] Furthermore, the above-described series of processes can be executed by hardware or software. In other words, the functional configurations in FIG. 3 and the like are merely examples and are not particularly limited. That is, it is sufficient that the information processing system is provided with a function capable of executing the above-mentioned series of processes as a whole, and the type of functional block used to realize this function is not limited to the example in Fig. 3 etc. Furthermore, the location of the functional block is not limited to the example in Fig. 3 etc. and may be arbitrary. Furthermore, one functional block may be configured as a single piece of hardware, may be configured as a single piece of software, or may be configured as a combination of both.

[0058] Also, for example, in the above embodiment, it has been described that one inference server 1, one learning server 2, one factory terminal 3, and only a plurality of imaging devices 4 are installed, but this is not limited thereto. In other words, the various hardware components and the number of users constituting the present system are not limited to the above-mentioned examples, and may be any number. Furthermore, the present system may be configured by including and combining not only the various hardware components described above, but also other hardware components, etc. Also, for example, the present system may be configured to execute various processes via the Web, etc., as in a so-called cloud-type system, and send the results of the processes to various hardware.

[0059] Furthermore, when the series of processes is executed by software, the program constituting the software is installed into a computer or the like from a network or a recording medium.

[0060] The computer may be a computer that is built into dedicated hardware, or a computer that is capable of executing various functions by installing various programs. That is, for example, any computer, any mobile terminal such as a smartphone, etc. may be freely adopted for the various hardware in the above-described embodiments.

[0061] In addition, the recording medium containing such a program may not only be constituted by a removable medium (not shown) provided separately from the device main body in order to provide the program to a user, etc., but may also be constituted by a recording medium, etc. that is provided to the user in a state in which it is pre-installed in the device main body.

[0062] In this specification, the steps of describing the program to be recorded on the recording medium may be executed in parallel or individually, as well as in chronological order. Also, some of the steps described in the above embodiment may be omitted.

[0063] In addition, in this specification, the term "system" refers to an overall device that is composed of a plurality of devices, a plurality of means, etc.

[0064] Even when these other embodiments are adopted, the effects of the present embodiment are exhibited. In addition, the present embodiment can be appropriately combined with other embodiments, and other embodiments can be appropriately combined with each other.

[0065] In other words, it is sufficient for an information processing device to which the present invention is applied to have the following configuration, and various embodiments can be adopted. That is, the first aspect of the present invention is A determination target image acquisition means (for example, a determination target image acquisition unit 60) that acquires an image including a determination target as a determination target image; A criterion acquisition means (e.g., a criterion acquisition unit 61) that acquires information regarding a criterion determined based on a score calculated by applying statistical processing to a learning image composed of a normal image; an abnormality detection means (e.g., a determination unit 62) that determines whether the target image is a normal image or an abnormal image based on the determination criterion; Any information processing device having the above will suffice.

[0066] In addition, in the first aspect, the anomaly detection means may determine whether the image to be judged is a normal image or an abnormal image based on a first threshold value determined by estimating a range of normality from a frequency distribution obtained from the learning image.

[0067] In addition, in the first aspect, the anomaly detection means may generate a pseudo-abnormal image, which is a pseudo-abnormal image, from a normal image, and determine whether the image to be judged is a normal image or an abnormal image based on a second threshold determined based on the normal image and the pseudo-abnormal image.

[0068] In addition, in the first aspect, when the result of the judgment by the anomaly detection means differs from the result of a judgment of the image to be judged by a human, the anomaly detection means may judge whether the image to be judged is a normal image or an abnormal image based on a third threshold value that is changed taking into account the result of the judgment by the human.

[0069] In addition, the second aspect of the present invention is A learning image acquisition means for acquiring a normal image as a learning image; a criterion generating means for generating a criterion for anomaly detection based on a calculated score by applying a statistical process to the learning images that are composed only of normal images and do not include any abnormal images; Any information processing device having the above will suffice.

[0070] In addition, the third aspect of the present invention is 1. A computer-implemented information processing method, comprising: a determination target image acquisition step of acquiring an image including a determination target as a determination target image; a criterion acquisition step of acquiring information about a criterion determined based on a score calculated by applying a statistical process to a learning image composed of normal images; an anomaly detection step of determining whether the target image is a normal image or an abnormal image based on the determination criterion; Any information processing method including the above will suffice.

[0071] Moreover, a fourth aspect of the present invention is On the computer, a determination target image acquisition step of acquiring an image including a determination target as a determination target image; a criterion acquisition step of acquiring information about a criterion determined based on a score calculated by applying a statistical process to a learning image composed of normal images; an anomaly detection step of determining whether the target image is a normal image or an abnormal image based on the determination criterion; Any program that executes a process including the above will suffice. [Explanation of symbols]

[0072] 1. Inference server 11. Control section 60: Judgment target image acquisition unit 61... Judgment criteria acquisition section 62...Judgment section 63...Judgment result management department 2. Learning Server 200...Control section 220....Learning image acquisition unit 221···Learning Department 222...Judgment criteria generation unit 3. Factory terminal 300...Control section 320...Image Management Section 321...Judgment result acquisition unit 4. Imaging device

Claims

1. a determination target image acquisition means for acquiring an image including a determination target as a determination target image; A criterion acquisition means for acquiring information regarding a criterion determined based on a score calculated by applying a statistical process to a learning image composed of normal images; an anomaly detection means for determining whether the image to be judged is a normal image or an abnormal image based on the judgment criterion; An information processing device comprising:

2. the anomaly detection means determines whether the target image is a normal image or an abnormal image based on a first threshold determined by estimating a range of normality from a frequency distribution obtained from the learning image; The information processing device according to claim 1 .

3. the abnormality detection means generates a pseudo-abnormal image, which is a pseudo-abnormal image, from a normal image, and judges whether the image to be judged is a normal image or an abnormal image based on a second threshold determined based on the normal image and the pseudo-abnormal image; The information processing device according to claim 1 .

4. When a result of the judgment by the anomaly detection means differs from a result of a judgment of the judgment target image made by a human, the anomaly detection means judges whether the judgment target image is a normal image or an abnormal image based on a third threshold value that has been changed taking into account the result of the judgment made by the human.

4. The information processing device according to claim 2 or 3.

5. A learning image acquisition means for acquiring a normal image as a learning image; a criterion generating means for generating a criterion for anomaly detection based on a calculated score by applying a statistical process to the learning images that are composed only of normal images and do not include any abnormal images; An information processing device comprising:

6. 1. A computer-implemented information processing method, comprising: a determination target image acquisition step of acquiring an image including a determination target as a determination target image; a criterion acquisition step of acquiring information regarding a criterion determined based on a score calculated by applying a statistical process to a learning image composed of normal images; an anomaly detection step of determining whether the target image is a normal image or an abnormal image based on the determination criterion; An information processing method comprising:

7. On the computer, a determination target image acquisition step of acquiring an image including a determination target as a determination target image; a criterion acquisition step of acquiring information regarding a criterion determined based on a score calculated by applying a statistical process to a learning image composed of normal images; an anomaly detection step of determining whether the target image is a normal image or an abnormal image based on the determination criterion; A program that executes a process including:

Citation Information

Patent Citations

  • Anomaly detection system and method

    JP2022085429A