Image automatic detection system and method thereof
By constructing an automatic image detection system, utilizing modules of cameras and processors, the system automatically identifies the normal and abnormal states of BIOS interfaces, solving the problems of misreading and missed detection caused by manual inspection, and achieving efficient automated detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TES TOUCH EMBEDDED SOLUTIONS (XIAMEN) CO LTD
- Filing Date
- 2025-01-23
- Publication Date
- 2026-07-24
AI Technical Summary
In existing technologies, manual inspection of BIOS interface information is prone to fatigue, leading to misreading and missed detections.
An automatic image detection system is adopted, which uses a camera to capture normal and abnormal images of the BIOS interface. Through the input module, image editing module, training module and detection module in the processor, an automatic image detection model is constructed to automatically identify the normal and abnormal BIOS interface.
It improves the accuracy and efficiency of BIOS interface detection, reduces misreading and missed detection, and realizes automated anomaly detection.
Smart Images
Figure CN122453689A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to an automatic detection system, and more particularly to an automatic image detection system. Background Technology
[0002] In a production line, there are multiple inspection stages. When the Basic Input / Output System (BIOS) is used as the interface for these inspection stages, each stage needs to be manually verified to ensure its correctness. However, manual inspection of the information displayed on the BIOS screen can lead to fatigue. For example, when manually inspecting the same interface, misreading information can occur.
[0003] Therefore, there is a real need to improve the existing technology. Summary of the Invention
[0004] One embodiment of this disclosure provides an automatic image detection system, including a camera and a processor. The camera is configured to capture multiple normal images, multiple abnormal images, and an image to be detected. The processor is communicatively connected to the camera and includes an input module, an image editing module, a training module, and a detection module. The input module is configured to receive the normal images, the abnormal images, and the image to be detected. The image editing module is configured to center and adjust the angles of the normal and abnormal images to reduce image noise. The training module is configured to train the adjusted normal and abnormal images using an anomaly detection algorithm to obtain an automatic image detection model. The detection module is configured to analyze the image to be detected using the automatic image detection model to determine whether the image to be detected is normal or abnormal.
[0005] In some implementations, these normal images, these abnormal images, and the images to be detected each contain images of the BIOS interface.
[0006] In some implementations, the image editing module is further configured to adjust the pixel size of these normal images, these abnormal images, and the image to be detected to be consistent.
[0007] In some implementations, the pixel size of these images is adjusted to 512×512 pixels.
[0008] In some implementations, the camera is further configured to shoot normal video, which is shot for multiple seconds at a frame rate of 10fps to 100fps and a resolution of 640×480 pixels to 61440×35640 pixels, and multiple frames are captured to form these normal images.
[0009] In some implementations, the camera is further configured to capture anomalous video, which is captured for multiple seconds at a frame rate of 10fps to 100fps and a resolution of 640×480 pixels to 61440×35640 pixels, and multiple frames are captured to form these anomalous images.
[0010] In some implementations, the automatic image inspection system is configured to allow users to inspect the images of the BIOS interfaces at multiple production stages during the production process to determine whether they are normal or abnormal.
[0011] In some implementations, the processor also includes an output module configured to automatically issue alerts or reminders about the results, enabling users to promptly correct these production processes.
[0012] In some implementations, anomaly detection algorithms include Reverse Distillation, Patchcore, EfficientAD, FastFlow, Padim, or combinations thereof.
[0013] Another embodiment of this disclosure provides a method for automatic image detection, comprising: capturing multiple normal images, multiple abnormal images, and an image to be detected using a camera; a processor including an input module, an image editing module, a training module, and a detection module, wherein: the input module receives the normal images, the abnormal images, and the image to be detected; the image editing module adjusts the normal images and the abnormal images to be centered and at the same angle to reduce image noise; the training module trains the adjusted normal images and the abnormal images using an anomaly detection algorithm to obtain an automatic image detection model; and the detection module analyzes the image to be detected using the automatic image detection model to identify whether the image to be detected is normal or abnormal.
[0014] In some implementations, these normal images, these abnormal images, and the images to be detected each contain images of the BIOS interface.
[0015] In some implementations, the method further includes using an image editing module to adjust the pixel size of these images to be consistent.
[0016] In some implementations, the pixel size of these images is adjusted to 512×512 pixels.
[0017] In some implementations, the step of shooting with a camera includes shooting normal video with the camera for multiple seconds at a frame rate of 10fps to 100fps and a resolution of 640×480 pixels to 61440×35640 pixels, and capturing multiple frames to form these normal images.
[0018] In some implementations, the camera shooting step includes shooting abnormal video with the camera for multiple seconds at a frame rate of 10fps to 100fps and a resolution of 640×480 pixels to 61440×35640 pixels, and capturing multiple frames to form these abnormal images.
[0019] In some implementations, the method also includes allowing users to inspect the images of BIOS interfaces at multiple production stages during the production process to determine whether they are normal or abnormal.
[0020] In some implementations, the processor also includes an output module that automatically issues alerts or reminders about the results, enabling users to promptly correct these production processes.
[0021] In some implementations, anomaly detection algorithms include Reverse Distillation, Patchcore, EfficientAD, FastFlow, Padim, or combinations thereof. Attached Figure Description
[0022] The various aspects of this disclosure will be most readily understood when read in conjunction with the accompanying drawings. It should be noted that, according to industry standard operating procedures, the various features may not be drawn to scale. In fact, for clarity of explanation, the dimensions of the various features can be arbitrarily increased or decreased. To make the above and other objects, features, advantages, and embodiments of this disclosure more apparent and understandable, the accompanying drawings are described below:
[0023] Figure 1 A schematic diagram illustrating some embodiments of the automatic image detection system disclosed herein;
[0024] Figure 2 A flowchart illustrating a method for training an automatic image detection model according to some embodiments disclosed herein is shown;
[0025] Figure 3 A flowchart illustrating some embodiments of the automatic image detection method disclosed herein is provided.
[0026] Figure 4 A bar chart illustrating the prediction accuracy of some of the embodiments disclosed herein.
[0027] [Symbol Explanation]
[0028] 100: Automatic Image Detection System
[0029] 110: Camera
[0030] 120: Processor
[0031] 121: Input Module
[0032] 122: Video Editing Module
[0033] 123: Training Module
[0034] 124: Detection Module
[0035] 125: Output Module
[0036] 200, 300: Method
[0037] S210, S220, S230, S240, S310, S320: Steps
[0038] 4: Figure Detailed Implementation
[0039] To make the description of this disclosure more detailed and complete, illustrative descriptions of embodiments and specific examples of this disclosure are provided below. However, these are not the only forms of implementing or utilizing the specific examples of this disclosure. The various embodiments disclosed below can be combined or substituted with each other where advantageously possible, and other embodiments can be added to one embodiment without further description or explanation. In the following description, many specific details will be set forth in detail to enable the reader to fully understand the following embodiments. However, the embodiments of this disclosure can also be practiced without such specific details.
[0040] Additionally, spatial relative terms, such as "down" and "up," are used to conveniently describe the relative relationship of one element or feature to other elements or features in the accompanying drawings. These spatial relative terms are intended to encompass different orientations of the device during use or operation, in addition to those shown in the drawings. The device may be positioned otherwise (e.g., rotated 90 degrees or otherwise), and the spatial relative descriptions used herein may be interpreted accordingly.
[0041] In this document, unless otherwise specified in the text, “a” and “the” may refer to one or more. It will be further understood that the terms “comprising,” “including,” “having,” and similar terms as used herein specify the features, regions, integers, steps, operations, elements, and / or components described herein, but do not exclude one or more other features, regions, integers, steps, operations, elements, components, and / or groups thereof described or additionally described herein.
[0042] Furthermore, when numbers or ranges of numbers are described using terms such as “about,” “approximately,” and the like, the term is intended to cover numbers within a reasonable range of variation that, as understood by those skilled in the art, inherently occurs during manufacturing. For example, based on known manufacturing tolerances associated with manufacturing characteristics (having properties associated with the number), a number or range of numbers encompasses a reasonable range including the described number, such as within + / –10% of the described number. Furthermore, references to numbers and / or letters may be repeated in various instances in this disclosure. This repetition is for simplicity and clarity and does not in itself indicate a relationship between the various embodiments and / or configurations discussed.
[0043] In this article, the term "image noise," also known as "noise" or "speculiarity," refers to random variations in brightness or color information within an image (not present in the object being photographed), typically manifested as electronic noise. It is generally generated by the camera's sensor and circuitry. Image noise is an undesirable byproduct of the image capture process, introducing errors and extraneous information into the image.
[0044] In this article, the term "frame rate" or "frame rate" refers to the speed at which a certain number of frames are displayed per second, and is therefore also referred to as frames per second (FPS), "frame rate" or "Hertz".
[0045] In this article, the terms "coupled" or "connected" can refer to two or more components making direct physical or electrical contact with each other, or making indirect physical or electrical contact with each other. "Coupled" or "connected" can also refer to two or more components operating or moving with each other.
[0046] The following examples and experimental cases illustrate the automatic image detection system and method disclosed herein in more detail. However, they are only illustrative and are not intended to limit the scope of the disclosure. The scope of protection of the disclosure shall be defined by the appended claims.
[0047] When using the Basic Input / Output System (BIOS) as the interface for inspection, each step needs to be manually verified to ensure its correctness. However, manual inspection of the information displayed on the BIOS screen can lead to fatigue. For example, manual inspection of the same interface can result in misreading certain parts. Furthermore, currently, the process relies solely on photographing / videotaping each stage of production for documentation, and manual verification of these stages can result in omissions.
[0048] This disclosure utilizes an automatic image detection system, combined with the powerful computing capabilities of GPUs, to train a suitable model based on real-world application scenarios, enabling the detection of production line-related anomalies. Unlike earlier manual detection or OCR-based recognition, this disclosure employs an anomaly detection algorithm combined with an artificial intelligence model and training methods. Based on a pre-collected dataset, appropriate cropping is performed, and an automatic image detection model is trained within a certain timeframe. Compared to existing technologies, this automatic image detection exhibits better generalization characteristics, capable of detecting blurry images with a high recognition rate.
[0049] The purpose of this disclosure of some embodiments is primarily to address the technical field of automatic image detection systems.
[0050] Figure 1 This is a schematic diagram of an automatic image detection system 100 according to some embodiments of this disclosure. For example... Figure 1 As illustrated, the automatic image detection system 100 includes a camera 110 and a processor 120. In some embodiments, the automatic image detection system 100 may include a storage device (not shown) and a communication chip (not shown). The storage device may be implemented as a memory, hard disk, flash drive, memory card, etc. In some embodiments, the automatic image detection system 100 is configured to allow users to inspect the images of the BIOS interfaces at multiple production stages during the production process to determine whether they are normal or abnormal.
[0051] Camera 110, also known as a video camera, is configured to capture multiple normal images, multiple abnormal images, and images to be detected. In some embodiments, camera 110 may include, but is not limited to, smart electronic devices such as mobile phones, computers, and tablets with photo and / or video recording functions. In some embodiments, the normal images, the abnormal images, and the images to be detected each include images from the BIOS interface. In some embodiments, camera 110 is further configured to capture normal video, which is captured for multiple seconds at a frame rate of 30fps and a resolution of 1920×1080 pixels, and multiple frames are captured to form multiple normal images. In some embodiments, camera 110 is further configured to capture abnormal video, which is captured for multiple seconds at a frame rate of 30fps and a resolution of 1920×1080 pixels, and multiple frames are captured to form multiple abnormal images. In some embodiments, the normal or abnormal video footage captured by camera 110 has a frame rate including but not limited to 10fps to 100fps and a resolution including but not limited to 640×480 pixels to 61440×35640 pixels for multiple seconds, such as DV (480×720), 720P, 1080P, 1K, 2K, 4K, 8K, 12K, 16K, 24K, 32K, and 64K pixels. In some embodiments, the frame rate is selected based on the number of images involved in the calculation. In some embodiments, the image resolution is selected based on the image quality of the images involved in the calculation.
[0052] Processor 120 is communicatively connected to camera 110. In some embodiments, processor 120 may be implemented as an integrated circuit such as a microcontroller, microprocessor, digital signal processor, application-specific integrated circuit (ASIC), logic circuit, or other similar elements or combinations thereof. The communication chip may be implemented as a global system for mobile communication (GSM), personal handyphone system (PHS), long term evolution system (LTE), worldwide interoperability for microwave access (WiMAX), wireless fidelity system (Wi-Fi), or Bluetooth transmission, etc. In some embodiments, processor 120 includes an input module 121, an image editing module 122, a training module 123, a detection module 124, and an output module 125.
[0053] Input module 121 is configured to accept these normal images, these abnormal images, and images to be detected.
[0054] The image editing module 122 is configured to center and adjust the angles of these normal images and these abnormal images to reduce image noise. In some embodiments, the image editing module 122 is further configured to adjust the pixel size of these images to be consistent. In some embodiments, the pixel size of these images is adjusted to 512×512 pixels, for example, from the original size of 1920×1080 pixels to 512×512. These adjusted pixel sizes are merely illustrative and can be determined according to the size of the image to be analyzed, and are not limited to those shown in this disclosure.
[0055] Training module 123 is configured to train the adjusted normal images and the abnormal images using an anomaly detection algorithm to obtain an automatic image detection model. In some embodiments, the anomaly detection algorithm includes Reverse Distillation, Patchcore, EfficientAD, FastFlow, Padim, or a combination thereof.
[0056] The detection module 124 is configured to analyze the image to be detected using an automatic image detection model to identify whether the image to be detected is normal or abnormal.
[0057] The output module 125 is configured to automatically issue alerts and reminders about the results, so that users can promptly correct these production processes.
[0058] To describe in detail how the automatic image detection system 100 operates, the following will be accompanied by... Figures 2 to 3 To explain. It should be understood that, in Figures 2 to 3 Unless otherwise specified, the order of the steps mentioned herein can be adjusted as needed, and they can be performed simultaneously or partially simultaneously. Additional steps can be added or some steps can be omitted.
[0059] Please continue reading. Figure 2 . Figure 2 The flowchart illustrates a method 200 for training an automatic image detection model according to some embodiments disclosed herein. In step S210, a camera 110 captures multiple normal images and multiple abnormal images. Next, in step S220, the input module 121 of the processor 120 receives these normal images and these abnormal images. Next, in step S230, the image editing module 122 of the processor 120 centers and adjusts the angles of these normal images and abnormal images to reduce noise. Next, in step S240, the training module 123 of the processor 120 trains the adjusted normal images and abnormal images using an anomaly detection algorithm to obtain an automatic image detection model.
[0060] In detail, this disclosure primarily involves first collecting a large amount of correct data (e.g., normal images), and then performing necessary cropping. Simultaneously, a suitable amount of anomalous data (e.g., anomalous images) is collected and categorized, with the aim of generating differences between the anomalous data and the correct data during training to enhance the effectiveness of the trained model.
[0061] In step S210, multiple normal images and multiple abnormal images are captured using camera 110. Specifically, to construct a dataset, for each normal and abnormal image, a video clip is recorded using camera 110 (e.g., a smartphone) at a playback rate of 30fps and a resolution of 1920×1080 pixels. During recording, camera 110 is moved to capture images of the BIOS interface displayed on the computer screen from multiple angles, even less than optimal angles, to introduce noise that would occur in real-world scenarios. On average, the normal images are captured for approximately 120 seconds, while the abnormal images are captured for approximately 30 seconds. Furthermore, to introduce additional noise, a wooden crate is placed on a carpet with many colored shapes. Therefore, as camera 110 is moved while positioned on the crate, the shapes of the noise in the background also change.
[0062] Next, in step S220, the normal images and abnormal images are received by the input module 121 of the processor 120. Specifically, the video captured by the camera 110 is transmitted to the computer, and the normal images and abnormal images are received by the input module 121 of the computer's processor 120.
[0063] Next, in step S230, the image editing module 122 of the processor 120 centers and adjusts the angles of these normal and abnormal images to reduce noise. Specifically, frames are extracted from the corresponding normal and abnormal images. Due to the noise added during shooting in various situations, each frame (image) is centered, cropped, and resized to 512×512 pixels.
[0064] Next, in step S240, the training module 123 of the processor 120 trains the adjusted normal images and abnormal images using an anomaly detection algorithm to obtain an automatic image detection model. Specifically, after collecting a sufficient dataset (normal images and abnormal images), anomaly detection algorithms are introduced, such as Reverse Distillation, Patchcore, EfficientAD, FastFlow, Padim, etc. In this embodiment, Reverse Distillation is mainly used.
[0065] The principle of Reverse Distillation lies in constructing corresponding Teacher (TS) patterns during the actual training of the model. T represents Teacher, and S represents Student. The knowledge learned by T is passed to S, allowing S to also acquire the ability to distinguish right from wrong. In traditional methods, because T and S use the same architecture, the resulting inference model may have ambiguity. The improved Reverse Distillation treats T as an encoder and S as a decoder, thus increasing the accuracy of the learned model.
[0066] This disclosure primarily utilizes Intel's Anomalib for implementation, focusing on BIOS interface checks. It can subsequently be applied to other aspects of production line inspection, such as packaging checks and arrangement checks.
[0067] Taking BIOS checks as an example, we first collected a large number of real-world BIOS interface images. The dataset was divided into two parts: correct BIOS interfaces and incorrect BIOS interfaces. For correct BIOS interfaces, we collected 3859 images. In one correct image, we intentionally added noise to simulate the complexity of a real-world environment (not shown). For incorrect BIOS interfaces, we prepared four sets of data (i.e., four error types: BIOS-Error 1, BIOS-Error 2, BIOS-Error 3, and BIOS-Error 4), with 449 images in each set, totaling 1796 images. These four sets of data mainly consist of abnormal data from the correct data, i.e., settings different from the normal data. During the image capture process, we deliberately introduced noise to enhance the fault tolerance of the generative model in experimental applications.
[0068] After training the automatic detection model with the above multiple normal and abnormal images as input images, the following results were produced.
[0069] Table 1
[0070]
[0071]
[0072] As shown in Table 1 above, 365 images were actually normal and were also predicted to be normal by the model (true positives); 897 images were actually abnormal and were also predicted to be abnormal by the model (true negatives); 21 images were actually normal but were predicted to be abnormal by the model (false negatives); and 2 images were actually abnormal but were predicted to be normal by the model (false positives). Therefore, it can be seen that the effectiveness of the disclosed automatic image detection model in anomaly detection (true negatives) is far higher than the results of false negatives and false positives.
[0073] In addition, the normal image group (BIOS-Correct) and the four abnormal image groups with different error states (BIOS-Error 1, BIOS-Error 2, BIOS-Error 3, BIOS-Error 4) were re-entered into the automatic image detection model established in this disclosure, and their corresponding accuracy rates were calculated.
[0074] Figure 4 A bar chart illustrating the prediction accuracy of some of the embodiments disclosed herein. Figure 4 .Depend on Figure 4 It can be seen that the accuracy rate of BIOS error group 1 is 0.65, the accuracy rate of BIOS error group 2 is 0.59, the accuracy rate of BIOS error group 3 is 0.8, and the accuracy rate of BIOS error group 4 is 0.82.
[0075] Please continue reading. Figure 3 . Figure 3 A flowchart illustrating an automatic image detection method 300 according to some embodiments disclosed herein is shown. In step S310, the detection module 124 of the processor 120 analyzes the image to be detected using an automatic image detection model to identify whether the image to be detected is normal or abnormal. Next, in step S320, the output module 125 of the processor 120 automatically issues a warning reminder based on the result, so that the user can promptly correct these production steps in the production process.
[0076] In detail, after training, the generated automatic image detection model can be directly applied to actual production. It analyzes the image data acquired by camera 110 and automatically distinguishes between normal and abnormal results. In step S310, camera 110 captures the image to be detected, and the detection module 124 of processor 120 analyzes the image using the automatic image detection model. For example, after obtaining the original image to be detected, it is converted into a heatmap and image segmentation is performed. Then, masking is performed to identify error regions by drawing image boxes. It is evident that the automatic image detection model disclosed herein can automatically compare normal and abnormal datasets, marking abnormal regions to identify abnormal phenomena.
[0077] Next, in step S320, the output module 125 of the processor 120 automatically sends out warnings and reminders to enable users to correct these production links in the production process in a timely manner.
[0078] Although the present disclosure has been described above with reference to embodiments, it is not intended to limit the present disclosure. Any person skilled in the art may make various modifications and refinements without departing from the spirit and scope of the present disclosure. Therefore, the scope of protection of the present disclosure shall be determined by the appended claims.
Claims
1. An automatic image detection system, characterized in that, Include: A camera configured to capture multiple normal images, multiple abnormal images, and one image to be detected; and A processor, communicatively connected to the camera, and comprising: An input module is configured to accept the plurality of normal images, the plurality of abnormal images, and the image to be detected; An image editing module is configured to center and adjust the angles of the multiple normal images and the multiple abnormal images to reduce image noise; A training module configured to train an automatic image detection model by using an anomaly detection algorithm to train the adjusted plurality of normal images and the plurality of abnormal images; and A detection module is configured to analyze the image to be detected using the automatic image detection model to determine whether the image to be detected is normal or abnormal.
2. The automatic image detection system as described in claim 1, characterized in that, The plurality of normal images, the plurality of abnormal images, and the image to be detected each contain an image of the BIOS interface.
3. The automatic image detection system as described in claim 1, characterized in that, The image editing module is further configured to adjust the pixel size of the plurality of normal images, the plurality of abnormal images and the image to be detected to be consistent.
4. The automatic image detection system as described in claim 3, characterized in that, The pixel size of the images containing the BIOS interface in the multiple normal images, the multiple abnormal images, and the image to be detected is adjusted to 512×512 pixels.
5. The automatic image detection system as described in claim 1, characterized in that, The camera is further configured to shoot a normal video at a frame rate of 10fps to 100fps and a resolution of 640×480 pixels to 61440×35640 pixels for multiple seconds, and capture multiple frames to form the multiple normal images.
6. The automatic image detection system as described in claim 1, characterized in that, The camera is further configured to capture an anomalous video, which is captured for multiple seconds at a frame rate of 10fps to 100fps and a resolution of 640×480 pixels to 61440×35640 pixels, and multiple frames are captured to form the multiple anomalous images.
7. The automatic image detection system as described in claim 1, characterized in that, This automatic image inspection system is configured to allow a user to inspect the images of the BIOS interfaces at multiple production stages during a production process to determine whether they are normal or abnormal.
8. The automatic image detection system as described in claim 7, characterized in that, The processor also includes an output module configured to automatically issue an alert about the result, so that the user can promptly correct the multiple production stages of the production process.
9. The automatic image detection system as described in claim 1, characterized in that, The anomaly detection algorithm includes Reverse Distillation, Patchcore, EfficientAD, FastFlow, Padim, or a combination thereof.
10. A method for automatic image detection, characterized in that, Include: A camera captures multiple normal images, multiple abnormal images, and one image to be detected. A processor includes an input module, an image editing module, a training module, and a detection module, wherein: The input module receives the multiple normal images, these abnormal images, and the image to be detected. The image editing module is used to center and adjust the angles of the multiple normal images and the multiple abnormal images to reduce image noise. The training module is used to train the adjusted multiple normal images and multiple abnormal images using an anomaly detection algorithm to obtain an automatic image detection model; and The detection module analyzes an image to be detected using the automatic image detection model to determine whether the image to be detected is normal or abnormal.
11. The method as described in claim 10, characterized in that, The plurality of normal images, the plurality of abnormal images, and the image to be detected each contain an image of the BIOS interface.
12. The method as described in claim 10, characterized in that, It also includes using the image editing module to adjust the pixel size of the multiple normal images, the multiple abnormal images, and the image to be detected to be consistent with the image containing the BIOS interface.
13. The method as described in claim 12, characterized in that, The pixel size of the images containing the BIOS interface in the multiple normal images, the multiple abnormal images, and the image to be detected is adjusted to 512×512 pixels.
14. The method as described in claim 10, characterized in that, The step of shooting with the camera includes shooting a normal video with the camera, the normal video being shot for multiple seconds at a frame rate of 10fps to 100fps and a resolution of 640×480 pixels to 61440×35640 pixels, and multiple frames being captured to become the multiple normal images.
15. The method as described in claim 10, characterized in that, The step of shooting with the camera includes shooting an abnormal video with the camera, the abnormal video being shot for multiple seconds at a frame rate of 10fps to 100fps and a resolution of 640×480 pixels to 61440×35640 pixels, and multiple frames being captured to become the multiple abnormal images.
16. The method as described in claim 10, characterized in that, It also includes images of the BIOS interface at multiple production stages to allow a user to check whether they are normal or abnormal during a production process.
17. The method as described in claim 16, characterized in that, The processor also includes an output module that automatically sends an alert about the result so that the user can promptly correct the multiple production stages of the production process.
18. The method as described in claim 10, characterized in that, The anomaly detection algorithm includes ReverseDistillation, Patchcore, EfficientAD, FastFlow, Padim, or a combination thereof.