Unknown industrial defect detection method, system and equipment and storage medium

Abstract

The invention provides an unknown industrial defect detection method, system and device and a storage medium, and relates to the technical field of industrial defect detection.The process of the method comprises the steps that an original industrial image is obtained and input into an industrial defect detection model for target detection, so that whether an unknown industrial defect exists on the surface of the original industrial image or not is judged; the industrial defect detection model comprises a target detection layer, and the target detection layer comprises a global pooling branch, a full connection branch and a result summarization branch; the global pooling branch obtains the similarity between the local feature of the original industrial image and the feature of the known industrial defect according to the similarity calculation; the full-connection branch calculates and obtains the local features of the original industrial image and the category confidence degree of the known industrial defect according to the placement degree; and the result summarization branch judges whether unknown industrial defects exist on the surface or not according to the feature similarity and the category confidence. According to the method, the problem of missing detection of unknown defects is successfully relieved in a manner of combining clustering analysis and target detection.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial defect detection, and in particular to an unknown industrial defect detection method, system, equipment and storage medium. Background Art

[0002] Industrial defect detection is a vital link in modern manufacturing. Its main purpose is to identify and locate various defects of products by automated means to ensure product quality and production efficiency. With the intensification of global market competition and the increasing demand of consumers for high-quality products, industrial defect detection has become a key means for enterprises to improve competitiveness, reduce costs and meet strict quality standards.

[0003] With the improvement of deep learning and big data analysis capabilities, artificial intelligence technology (AI technology) has shown great potential in the field of industrial defect detection. First, the artificial intelligence system realizes automatic detection of product appearance through visual recognition. That is, using deep learning algorithms such as convolutional neural networks (CNN networks), artificial intelligence systems can take high-speed photos of products on the production line and accurately identify various appearance defects. Compared with traditional manual visual inspection, artificial intelligence systems can achieve continuous and uninterrupted automatic inspection, greatly improving production efficiency and product quality. Secondly, defect classification technology based on machine learning models enables artificial intelligence systems to distinguish various types of defects and provide timely feedback and adjustment suggestions for subsequent production processes. By training a large number of defect samples, artificial intelligence systems can gradually improve recognition accuracy and respond quickly to new defects, thereby minimizing the defective rate.

[0004] In general, the application of artificial intelligence systems in industrial defect detection continues to expand and deepen, bringing more efficient and accurate quality control methods to the manufacturing industry. However, industrial inspections are still difficult to conduct reasonable and effective detection of unseen defects. Such defects are often mistakenly detected by the model as known defects, which in turn affects factory production. Summary of the invention

[0005] In order to solve the problem that unknown industrial defects are easily misdetected during existing industrial defect detection, an embodiment of the present invention provides an unknown industrial defect detection method, system, device and storage medium.

[0006] In a first aspect, an embodiment of the present invention provides a method for detecting unknown industrial defects, the method comprising the following steps: Acquire original industrial images and preprocess the original industrial images; The pre-processed original industrial image is input into the industrial defect detection model for target detection to determine whether there are unknown industrial defects on the surface of the original industrial image; The industrial defect detection model includes a target detection layer, wherein the target detection layer includes a global pooling branch, a fully connected branch, and a result summary branch; The global pooling branch obtains the feature similarity between the local features of the original industrial image and the known industrial defects according to the similarity calculation; The fully connected branch obtains the category confidence of the local features of the original industrial image and the known industrial defects according to the confidence calculation; The result summary branch determines whether there are unknown industrial defects on the surface according to the feature similarity and category confidence between the local features of the original industrial image and the known industrial defects.

[0007] In the above embodiment, the present invention improves the structure and function of the industrial defect detection model so that the improved industrial defect detection model can detect unknown industrial defects. By combining cluster analysis and target detection, the problem of missed detection of unknown defects is successfully alleviated.

[0008] As some optional embodiments of the present application, the industrial defect detection model includes an input layer, a convolution layer, a pooling layer, a target detection layer and an output layer arranged in series, and the target detection layer includes a global pooling branch, a fully connected branch and a result summary branch arranged in parallel with the global pooling branch and the fully connected branch.

[0009] In the above embodiment, the industrial defect detection model of the present invention is improved on the basis of the traditional CNN model, and the fully connected layer of the original CNN model is improved into a target detection layer, so that the industrial defect detection model has the ability to detect unknown industrial defects.

[0010] As some optional implementations of the present application, before the pre-processed original industrial image is input into the industrial defect detection model for target detection, the industrial defect detection model needs to be trained.

[0011] In the above embodiment, the present invention trains the industrial defect detection model so that the industrial defect detection model obtains global pooling features of all known defect categories to facilitate the subsequent calculation of feature similarity and category confidence.

[0012] As some optional implementation methods of the present application, the process of model training for the industrial defect detection model is as follows: Obtain original industrial images, preprocess them, and label their defect categories; The original industrial images with defect category annotations are input into the industrial defect detection model for model training to obtain the global pooling features of all known defect categories.

[0013] In the above embodiment, the present invention improves the structure and function of the industrial defect detection model so that the improved industrial defect detection model can detect unknown industrial defects. By combining cluster analysis and target detection, the problem of missed detection of unknown defects is successfully alleviated, which brings higher accuracy and reliability to defect detection in industrial production, and has important practical application value and promotion prospects.

[0014] As some optional implementations of the present application, the process of obtaining the feature similarity between the local features of the original industrial image and the known industrial defects according to the similarity calculation by the global pooling branch is as follows: Obtain local pooling features of original industrial images; The similarities between the local pooled features of the original industrial image and the global pooled features of all known defect categories are calculated respectively to obtain the feature similarity of the local pooled features of the original industrial image corresponding to each known defect category.

[0015] In the above embodiment, the present invention obtains an indicator parameter for unknown defect determination by calculating the similarity of feature communities.

[0016] As some optional implementations of the present application, the process of the fully connected branch obtaining the category confidence of the local features of the original industrial image and the known defect category according to the confidence calculation is as follows: Obtain local pooling features of original industrial images; Perform feature classification on the local pooled features of the original industrial image to obtain defect classification results; The defect classification results are normalized to obtain the category confidence of each known defect category corresponding to the local pooled features of the original industrial image.

[0017] In the above embodiment, the present invention obtains another index parameter for unknown defect determination by means of variance elimination.

[0018] As some optional implementations of the present application, the process of the result summary branch determining whether there are unknown industrial defects on the surface according to the feature similarity and category confidence between the local features of the original industrial image and the known industrial defects is as follows: Comprehensively detect the feature similarity and category confidence between the local features of the original industrial image and the known industrial defects; Based on the similarity threshold, the local pooling features of the original industrial image are matched and checked against the similarity values ​​of each known defect category, and the corresponding similarity determination results are output according to the matching and verification results; Based on the confidence threshold, the variance check is performed on the category confidence of each known defect category corresponding to the local pooling feature of the original industrial image, and the corresponding confidence judgment result is output according to the variance check result; Based on the comprehensive judgment rule, the similarity judgment result and the confidence judgment result are comprehensively judged, and whether there are unknown industrial defects on the surface of the original industrial image is output according to the comprehensive judgment result.

[0019] In the above-mentioned embodiment, the comprehensive verification method of the present invention is used to determine unknown defects, so that the defect determination accuracy is higher and the occurrence of false detection can be effectively prevented.

[0020] In a second aspect, the present invention provides an unknown industrial defect detection system, the system comprising: An image acquisition unit, the image acquisition unit is used to acquire an original industrial image and preprocess the original industrial image; A target detection unit, wherein the target detection unit is used to input the preprocessed original industrial image into the industrial defect detection model for target detection to determine whether there are unknown industrial defects on the surface of the original industrial image; The industrial defect detection model includes a target detection layer, wherein the target detection layer includes a global pooling branch, a fully connected branch, and a result summary branch; The global pooling branch obtains the feature similarity between the local features of the original industrial image and the known industrial defects according to the similarity calculation; The fully connected branch obtains the category confidence of the local features of the original industrial image and the known industrial defects according to the confidence calculation; The result summary branch determines whether there are unknown industrial defects on the surface according to the feature similarity and category confidence between the local features of the original industrial image and the known industrial defects.

[0021] In a third aspect, the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the unknown industrial defect detection method.

[0022] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the unknown industrial defect detection method when executed by a processor.

[0023] The beneficial effects of the present invention are as follows: The present invention improves the structure and function of the industrial defect detection model so that the improved industrial defect detection model can detect unknown industrial defects. By combining cluster analysis and target detection, the problem of missed detection of unknown defects is successfully alleviated, which brings higher accuracy and reliability to defect detection in industrial production, and has important practical application value and promotion prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 is a schematic diagram of the computer device structure of the hardware operating environment described in the embodiment of the present invention; Figure 2 is a flow chart of the unknown industrial defect detection method according to an embodiment of the present invention; Figure 3 It is a schematic diagram of the structure of the target detection layer described in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0027] In order to solve the problem of misdetection of unknown industrial defects in existing industrial defect detection, the embodiment of the present invention provides an unknown industrial defect detection method, system, device and storage medium. Before introducing the specific technical solution of the present application, the hardware operating environment involved in the embodiment of the present application is first introduced.

[0028] See also Figure 1 , Figure 1 The figure is a schematic diagram of the computer device structure of the hardware operating environment involved in the embodiment of the present invention.

[0029] like Figure 1 As shown, the computer device may include: a processor, such as a central processing unit (CPU), a communication bus, a user interface, a network interface, and a memory. Among them, the communication bus is used to realize the connection and communication between these components. The user interface may include a display screen (Display), an input unit such as a keyboard (Keyboard), and optionally, the user interface may also include a standard wired interface and a wireless interface. The network interface may be a wired interface or a wireless interface (such as a wireless fidelity interface). The memory may be a high-speed random access memory (Random Access Memory, RAM) or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory. The memory may optionally be a storage device independent of the aforementioned processor.

[0030] Those skilled in the art will understand that Figure 1The structure shown in the figure does not constitute a limitation on the computer device, and may include more or less components than shown in the figure, or combine some components, or arrange the components differently.

[0031] like Figure 1 As shown, the memory as a storage medium may include an operating system module, a network communication module, a user interface module and a data storage module.

[0032] exist Figure 1 In the computer device shown, the network interface is mainly used for data communication with the network server; the user interface is mainly used for data interaction with the user; the processor and memory in the computer device of the present application can be set in the computer device, and the computer device calls the unknown industrial defect detection system stored in the data storage module through the processor, and executes the unknown industrial defect detection method provided in the embodiment of the present application.

[0033] Based on the hardware environment of the aforementioned embodiment, the embodiment of the present application provides a panel zero-sample defect detection method, see Figure 2 , Figure 2 The flowchart of the unknown industrial defect detection method is as follows: (1) Obtaining an original industrial image to be inspected and preprocessing the original industrial image, wherein the preprocessing method of the original image includes but is not limited to filtering denoising, image enhancement, etc.

[0034] In an embodiment of the present invention, before the pre-processed original industrial image is input into the industrial defect detection model for target detection, the industrial defect detection model needs to be trained.

[0035] In an embodiment of the present invention, the industrial defect detection model includes an input layer, a convolution layer, a pooling layer, a target detection layer and an output layer arranged in series, and the target detection layer includes a global pooling branch, a fully connected branch arranged in parallel, and a result summary branch arranged in series with the global pooling branch and the fully connected branch.

[0036] Among them, the input layer is responsible for receiving the original industrial image, the convolution layer is responsible for extracting the local features of the original industrial image to generate a feature map; the pooling layer is responsible for reducing the spatial dimension of the feature map; the target detection layer is responsible for extracting the global pooling features of the known defect categories from the feature map during the model training stage, and in order to reduce the amount of calculation for subsequent similarity calculations, the embodiment of the present invention uses the centroid features of the community of each defect category to replace the information of the entire category community, wherein the centroid features of the community can be understood as the distribution center of the defect. During the model inference stage, the target detection layer is responsible for extracting the local pooling features of the original industrial image to be detected from the feature map, calculating the feature similarity and category confidence of the local pooling features and the known defect categories, and determining whether there are unknown industrial defects on the surface of the original industrial image based on the judgment decision; the output layer is responsible for outputting the defect judgment result.

[0037] Specifically, the process of model training for the industrial defect detection model is as follows: (1.1) Obtaining historical original industrial images, and preprocessing and labeling defect categories of the original industrial images; wherein the defect category labeling is labeling defect locations and defect categories of defects on the surface of the original industrial images.

[0038] (1.2) The original industrial images with defect category annotations are input into the industrial defect detection model for model training to obtain the global pooling features of all known defect categories.

[0039] (2) The preprocessed original industrial image is input into the industrial defect detection model for target detection to determine whether there are unknown industrial defects on the surface of the original industrial image.

[0040] First, the global pooling branch obtains the feature similarity between the local features of the original industrial image and the known industrial defects based on similarity calculation.

[0041] Then, the fully connected branch obtains the category confidence of the local features of the original industrial image and the known industrial defects according to the confidence calculation.

[0042] Finally, the result summary branch determines whether there are unknown industrial defects on the surface based on the feature similarity and category confidence between the local features of the original industrial image and the known industrial defects. Figure 3 , Figure 3 Schematic diagram of the structure of the target detection layer.

[0043] In the embodiment of the present invention, the process of the global pooling branch obtaining the feature similarity between the local features of the original industrial image and the known industrial defects according to the similarity calculation is as follows: (2.11) Obtain local pooling features of the original industrial image.

[0044] (2.12) Calculate the similarity between the local pooled features of the original industrial image and the global pooled features of all known defect categories respectively to obtain the feature similarity of the local pooled features of the original industrial image corresponding to each known defect category.

[0045] Specifically, the feature similarity calculation method includes but is not limited to the Manhattan distance formula, the cosine similarity formula, the Pearson correlation coefficient formula, etc.; preferably, the embodiment of the present invention uses the cosine similarity calculation formula.

[0046] Wherein, the cosine similarity calculation formula is:

[0047] Among them, A and B represent two eigenvectors respectively, i represents the feature point of the eigenvector, n represents the number of feature points of the eigenvector, and the chord similarity ranges from -1 to 1, where -1 represents completely opposite and 1 represents completely identical.

[0048] In the embodiment of the present invention, the process of the fully connected branch obtaining the category confidence of the local features of the original industrial image and the known defect category according to the confidence degree calculation is as follows: (2.21) Obtain the local pooling features of the original industrial image.

[0049] (2.22) Feature classification is performed on the local pooled features of the original industrial image to obtain the defect classification results.

[0050] (2.23) The defect classification results are normalized to obtain the category confidence of each known defect category corresponding to the local pooled features of the original industrial image. Specifically, since the sum of all predicted categories is 1 after normalization, once the predicted score of a certain category is much higher than that of other categories, it can be considered that the category is a known category. On the contrary, if the overall score appears to be relatively average, it means that the defects in the original industrial image may not be similar to any known category and are unknown defect categories.

[0051] In the embodiment of the present invention, the process of the result summary branch determining whether there are unknown industrial defects on the surface according to the feature similarity and category confidence between the local features of the original industrial image and the known industrial defects is as follows: (2.31) Comprehensively detect the feature similarity and category confidence between the local features of the original industrial image and the known industrial defects.

[0052] (2.32) Based on the similarity threshold, the similarity values ​​of the local pooled features of the original industrial image corresponding to each known defect category are matched and verified, and the corresponding similarity judgment results are output according to the matching verification results.

[0053] (2.33) Based on the confidence threshold, the variance of the category confidence of each known defect category corresponding to the local pooled features of the original industrial image is checked, and the corresponding confidence judgment result is output according to the variance check result.

[0054] (2.34) Based on the comprehensive judgment rule, the similarity judgment result and the confidence judgment result are comprehensively judged, and whether there are unknown industrial defects on the surface of the original industrial image is output according to the comprehensive judgment result.

[0055] Specifically, the process of comprehensive detection of the feature similarity and category confidence of the local features of the original industrial image and the known industrial defects is as follows: ① If the local pooling features of the original industrial image correspond to the similarity threshold of each known defect category, the output similarity judgment result is 1, otherwise, the output similarity judgment result is 0.

[0056] ② If the variance of the category confidence of each known defect category corresponding to the local pooling feature of the original industrial image is less than the confidence threshold, the output confidence judgment result is 1, otherwise, the output confidence judgment result is 0.

[0057] ③ If the similarity judgment result and the confidence judgment result are both 1, it is determined that the original industrial image includes unknown defect categories. Otherwise, it is determined that the original industrial image does not include unknown defect categories.

[0058] In summary, the embodiments of the present invention improve the structure and function of the industrial defect detection model so that the improved industrial defect detection model can detect unknown industrial defects. By combining cluster analysis and target detection, the problem of missed detection of unknown defects is successfully alleviated, which brings higher accuracy and reliability to defect detection in industrial production, and has important practical application value and promotion prospects.

[0059] In addition, in one embodiment, based on the same inventive concept as the above embodiment, the embodiment of the present invention provides an unknown industrial defect detection system, the system corresponds one-to-one with the method of embodiment 1, and the system includes: An image acquisition unit, the image acquisition unit is used to acquire an original industrial image and preprocess the original industrial image; A target detection unit, wherein the target detection unit is used to input the preprocessed original industrial image into the industrial defect detection model for target detection to determine whether there are unknown industrial defects on the surface of the original industrial image; The industrial defect detection model includes a target detection layer, wherein the target detection layer includes a global pooling branch, a fully connected branch, and a result summary branch; The global pooling branch obtains the feature similarity between the local features of the original industrial image and the known industrial defects according to the similarity calculation; The fully connected branch obtains the category confidence of the local features of the original industrial image and the known industrial defects according to the confidence calculation; The result summary branch determines whether there are unknown industrial defects on the surface according to the feature similarity and category confidence between the local features of the original industrial image and the known industrial defects.

[0060] It should be noted that each unit in the unknown industrial defect detection system in this embodiment corresponds one-to-one to each step in the unknown industrial defect detection method in the aforementioned embodiment. Therefore, the specific implementation method and technical effects achieved in this embodiment can refer to the implementation method of the aforementioned unknown industrial defect detection method, which will not be repeated here.

[0061] In addition, in one embodiment, the present application also provides a computer device, which includes a processor, a memory, and a computer program stored in the memory, and the computer program implements the method in the aforementioned embodiment when executed by the processor.

[0062] In addition, in one embodiment, the present application further provides a computer storage medium, on which a computer program is stored, and the computer program implements the method in the aforementioned embodiment when executed by a processor.

[0063] In some embodiments, the computer readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or various devices including one or any combination of the above memories. The computer may be various computing devices including intelligent terminals and servers.

[0064] In some embodiments, executable instructions may be in the form of a program, software, software module, script or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a stand-alone program or as a module, component, subroutine or other unit suitable for use in a computing environment.

[0065] As an example, executable instructions may, but need not, correspond to a file in a file system, may be stored as part of a file storing other programs or data, such as in one or more scripts in a HyperText Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files storing one or more modules, subroutines, or code portions).

[0066] By way of example, executable instructions may be deployed to be executed on one computing device, or on multiple computing devices located at one site, or on multiple computing devices distributed across multiple sites and interconnected by a communication network.

[0067] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0068] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0069] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory / random access memory, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a multimedia terminal device (which can be a mobile phone, a computer, a television receiver, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0070] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for detecting unknown industrial defects, characterized in that: The method comprises the following steps: Acquire original industrial images and preprocess the original industrial images; The pre-processed original industrial image is input into the industrial defect detection model for target detection to determine whether there are unknown industrial defects on the surface of the original industrial image; The industrial defect detection model includes a target detection layer, wherein the target detection layer includes a global pooling branch, a fully connected branch, and a result summary branch; The global pooling branch obtains the feature similarity between the local features of the original industrial image and the known industrial defects according to the similarity calculation; The fully connected branch obtains the category confidence of the local features of the original industrial image and the known industrial defects according to the confidence calculation; The result summary branch determines whether there are unknown industrial defects on the surface according to the feature similarity and category confidence between the local features of the original industrial image and the known industrial defects.

2. The unknown industrial defect detection method according to claim 1, characterized in that: The industrial defect detection model includes an input layer, a convolution layer, a pooling layer, a target detection layer and an output layer arranged in series, and the target detection layer includes a global pooling branch, a fully connected branch arranged in parallel, and a result summary branch arranged in series with the global pooling branch and the fully connected branch.

3. The unknown industrial defect detection method according to claim 1, characterized in that: Before inputting the preprocessed original industrial image into the industrial defect detection model for target detection, the industrial defect detection model needs to be trained.

4. The unknown industrial defect detection method according to claim 3, characterized in that: The process of model training for industrial defect detection models is as follows: Obtain original industrial images, preprocess them, and label their defect categories; The original industrial images with defect category annotations are input into the industrial defect detection model for model training to obtain the global pooling features of all known defect categories.

5. The unknown industrial defect detection method according to claim 4, characterized in that: The process of the global pooling branch obtaining the similarity between the local features of the original industrial image and the features of the known industrial defects based on similarity calculation is as follows: Obtain local pooling features of original industrial images; The similarities between the local pooled features of the original industrial image and the global pooled features of all known defect categories are calculated respectively to obtain the feature similarity of the local pooled features of the original industrial image corresponding to each known defect category.

6. The unknown industrial defect detection method according to claim 4, characterized in that: The process of the fully connected branch obtaining the category confidence of the local features of the original industrial image and the known defect categories according to the confidence degree calculation is as follows: Obtain local pooling features of original industrial images; Perform feature classification on the local pooled features of the original industrial image to obtain defect classification results; The defect classification results are normalized to obtain the category confidence of each known defect category corresponding to the local pooled features of the original industrial image.

7. The unknown industrial defect detection method according to claim 1, characterized in that: The process of the result summary branch determining whether there are unknown industrial defects on the surface according to the feature similarity and category confidence between the local features of the original industrial image and the known industrial defects is as follows: Comprehensively detect the feature similarity and category confidence between the local features of the original industrial image and the known industrial defects; Based on the similarity threshold, the local pooling features of the original industrial image are matched and checked against the similarity values ​​of each known defect category, and the corresponding similarity determination results are output according to the matching and verification results; Based on the confidence threshold, the variance check is performed on the category confidence of each known defect category corresponding to the local pooling feature of the original industrial image, and the corresponding confidence judgment result is output according to the variance check result; Based on the comprehensive judgment rule, the similarity judgment result and the confidence judgment result are comprehensively judged, and whether there are unknown industrial defects on the surface of the original industrial image is output according to the comprehensive judgment result.

8. An unknown industrial defect detection system, characterized in that: The system comprises: An image acquisition unit, the image acquisition unit is used to acquire an original industrial image and preprocess the original industrial image; A target detection unit, wherein the target detection unit is used to input the preprocessed original industrial image into the industrial defect detection model for target detection to determine whether there are unknown industrial defects on the surface of the original industrial image; The industrial defect detection model includes a target detection layer, wherein the target detection layer includes a global pooling branch, a fully connected branch, and a result summary branch; The global pooling branch obtains the feature similarity between the local features of the original industrial image and the known industrial defects according to the similarity calculation; The fully connected branch obtains the category confidence of the local features of the original industrial image and the known industrial defects according to the confidence calculation; The result summary branch determines whether there are unknown industrial defects on the surface according to the feature similarity and category confidence between the local features of the original industrial image and the known industrial defects.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for detecting unknown industrial defects described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for detecting unknown industrial defects described in any one of claims 1 to 7 is implemented.

Images