Real-time conductor surface defect detection method and system

Abstract

The invention relates to the technical field of conductor defect detection, and particularly provides a conductor surface defect real-time detection method and system, and the method comprises the steps: obtaining the transparency information of an insulating layer according to the optical characteristics of an optical image of which the optical wavelength is a preset wavelength, determining parameters of a region identification algorithm and parameters of a first defect detection algorithm according to the transparency information of the insulating layer; performing surface printing mark identification on the optical image of which the optical wavelength is within the preset wave band by using a region identification algorithm to obtain a surface printing mark region; performing feature extraction on areas except the surface printing identification areas in all the optical images to obtain a first feature map reflecting information of different layers; performing defect identification on the first feature map by using a first defect detection algorithm, and then determining the position of the surface defect according to the feature difference of the same surface defect in different optical images; the method can effectively improve the accuracy, reliability and robustness of defect detection.

Description

Technical Field

[0001] This application relates to the technical field of conductor defect detection, and more specifically, to a method and system for real-time detection of conductor surface defects. Background Art

[0002] When producing semi-transparent insulating conductors (such as insulated wires with semi-transparent insulating layers), the integrity and consistency of the insulating layer are directly related to the electrical performance and safety of the product. Existing technologies use automated inspection systems based on machine vision to detect defects in semi-transparent insulating conductors.

[0003] However, insulating layer wires with semi-transparent characteristics pose many challenges to existing automated inspection systems. First, the semi-transparency of the insulating layer material enables the camera to capture not only the features on the outer surface of the insulating layer but also, to a certain extent, the situation inside the insulating layer and even the conductor surface when collecting images. This means that the inspection system needs to simultaneously focus on surface defects of the insulating layer (such as scratches or surface roughness), internal defects of the insulating layer (such as bubbles or impurities), and abnormalities on the conductor surface that may be revealed through the insulating layer. These defects at different levels may overlap or interfere with each other in the image, making it difficult to accurately distinguish them. That is, existing automated inspection systems have problems with insufficient accuracy and reliability in defect detection of semi-transparent insulating conductors.

[0004] Second, the transparency of the insulating layer material may fluctuate under the influence of factors such as raw material batch differences, extrusion temperature, and cooling rate. Changes in the transparency of the insulating layer directly affect the visibility of the underlying conductor or internal features. Existing machine vision algorithms usually use fixed parameters for defect detection, and fixed-parameter defect detection is difficult to adapt to the impact brought about by such transparency fluctuations, resulting in a decline in the robustness and accuracy of defect detection. For example, when the transparency is high, internal and conductor features are clearer, but it is easier to misjudge normal textures on the conductor surface as defects; when the transparency is low, internal and conductor defects may become blurred or even invisible, leading to missed detections.

[0005] In addition, in order to identify product information, the outer surface of the insulating layer is usually marked by inkjet printing or laser printing (such as model, specification, production date, etc.). These printed marks interfere with defect detection. Therefore, existing technologies also have problems where printed marks are misjudged as defects due to interference with defect detection. In response to the above problems, there is currently no effective technical solution. It should be noted that the above information disclosed in this section is only used to understand the background of the inventive concept of the present invention and may therefore include information that does not constitute prior art. Summary of the Invention

[0006] The purpose of this application is to provide a real-time detection method and system for conductor surface defects, which can effectively improve the accuracy, reliability, and robustness of defect detection.

[0007] In a first aspect, this application provides a real-time detection method for conductor surface defects, which is used to detect surface defects of a semi-transparent insulating conductor, and includes the following steps: S1. Obtain a multi-spectral image sequence of the semi-transparent insulating conductor, where the multi-spectral image sequence includes optical images with different optical wavelengths; S2. Obtain the transparency information of the insulating layer according to the optical characteristics of the optical image with a preset optical wavelength, and then determine the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer; S3. Use the region recognition algorithm to identify surface printing marks in the optical images with optical wavelengths within a preset band to obtain the surface printing mark area; S4. Extract features from the areas in all optical images except the surface printing mark area to obtain a first feature map reflecting information at different levels; S5. Use the first defect detection algorithm to identify defects in the first feature map, and then when there are surface defects, determine the position of the surface defect according to the feature differences of the same surface defect in different optical images.

[0008] The real-time detection method for conductor surface defects provided by this application can make the region recognition algorithm and the first defect detection algorithm applicable to the current transparency of the insulating layer by determining the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer. Therefore, this application can effectively solve the problem that the robustness and accuracy of defect detection decrease due to the difficulty of fixed-parameter defect detection in adapting to the influence brought by this transparency fluctuation. And because this application can eliminate the interference of printing marks on defect detection by first using the region recognition algorithm to identify the surface printing mark area and then detecting defects in the areas in the optical images except the surface printing mark area, this application can effectively solve the problem that printing marks interfere with defect detection and lead to misjudging printing marks as defects. Moreover, because this application can realize defect recognition and positioning at different levels by extracting multi-spectral features and defect recognition from non-printing mark areas and using the feature differences of the same surface defect in different optical images to determine its position, this application can also effectively solve the problem that the accuracy and reliability of defect detection of semi-transparent insulating conductors are insufficient due to the possible superposition or interference of defects at different levels in the image, making it difficult to accurately distinguish. That is, this application can effectively improve the accuracy, reliability, and robustness of defect detection.

[0009] Second aspect, the present application also provides a real-time conductor surface defect detection system for detecting surface defects of a semi-transparent insulating conductor, which includes: An image sequence acquisition module for acquiring a multi-spectral image sequence of the semi-transparent insulating conductor, the multi-spectral image sequence including optical images of different optical wavelengths; An algorithm parameter confirmation module for obtaining the transparency information of the insulating layer according to the optical characteristics of the optical image with a preset optical wavelength, and then determining the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer; A printed mark area recognition module for using the region recognition algorithm to recognize the surface printed marks of the optical image with the optical wavelength within the preset band to obtain the surface printed mark area; A feature extraction module for extracting features from the areas other than the surface printed mark area in all optical images to obtain a first feature map reflecting information of different layers; A defect detection module for using the first defect detection algorithm to identify defects in the first feature map, and then when there are surface defects, determining the position of the surface defect according to the feature differences of the same surface defect in different optical images.

[0010] The real-time conductor surface defect detection system provided by the present application can make the region recognition algorithm and the first defect detection algorithm applicable to the current transparency of the insulating layer by determining the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer. Therefore, the present application can effectively solve the problem that the robustness and accuracy of defect detection decrease due to the difficulty of fixed-parameter defect detection in adapting to the influence brought by this transparency fluctuation. And because the present application can eliminate the interference of printed marks on defect detection by first using the region recognition algorithm to recognize the surface printed mark area and then detecting defects in the areas other than the surface printed mark area in the optical image, the present application can effectively solve the problem that printed marks interfere with defect detection and lead to misjudging printed marks as defects. Moreover, because the present application can realize defect recognition and positioning of different layers by extracting multi-spectral features and defect recognition of non-printed mark areas and using the feature differences of the same surface defect in different optical images to determine its position, the present application can also effectively solve the problem that the accuracy and reliability of defect detection of semi-transparent insulating conductors are insufficient due to the possible superposition or interference of defects in different layers in the image and the difficulty in accurately distinguishing them. That is, the present application can effectively improve the accuracy, reliability and robustness of defect detection.

[0011] As can be seen from the above, a real-time detection method and system for conductor surface defects provided by this application can make the region recognition algorithm and the first defect detection algorithm applicable to the current transparency of the insulating layer by determining the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer. Therefore, this application can effectively solve the problem that the robustness and accuracy of defect detection decrease due to the difficulty of fixed-parameter defect detection in adapting to the impact brought by such transparency fluctuations. Since this application can eliminate the interference of printing marks on defect detection by first using the region recognition algorithm to identify the surface printing identification region and then detecting defects in the region other than the surface printing identification region in the optical image, this application can effectively solve the problem that printing marks interfere with defect detection and lead to misjudging printing marks as defects. And since this application can achieve defect recognition and positioning at different levels by performing multi-spectral feature extraction and defect recognition on the non-printing identification region and determining its position by using the feature differences of the same surface defect in different optical images, this application can also effectively solve the problem that the accuracy and reliability of defect detection of semi-transparent insulating conductors are insufficient due to the possible superposition or interference of defects at different levels in the image, making it difficult to accurately distinguish. That is, this application can effectively improve the accuracy, reliability, and robustness of defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a flowchart of a real-time detection method for conductor surface defects provided by an embodiment of this application.

[0013] Figure 2 It is a schematic structural diagram of a real-time detection system for conductor surface defects provided by an embodiment of this application.

[0014] Reference numerals: 1, image sequence acquisition module; 2, algorithm parameter confirmation module; 3, printing identification region recognition module; 4, feature extraction module; 5, defect detection module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] Next, the technical solutions in the embodiments of this application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. The components of the embodiments of this application usually described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the drawings is not intended to limit the scope of the claimed application, but merely represents the selected embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of protection of this application.

[0016] It should be noted that: Similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present application, terms such as "first" and "second" are only used for distinguishing descriptions and cannot be construed as indicating or implying relative importance.

[0017] When traditional existing automated detection systems detect defects in semi-transparent insulating conductors, there are problems of insufficient accuracy, robustness, and reliability in defect detection of semi-transparent insulating conductors. The semi-transparency of the insulating layer material makes the images captured by the camera contain information on the outer surface of the insulating layer, the inside of the insulating layer, and the surface of the conductor. Defects at different levels are superimposed or interfered with each other in the image, making it difficult to accurately distinguish. In addition, the transparency of the insulating layer material may fluctuate, and existing machine vision algorithms usually use fixed parameters for defect detection, making it difficult to adapt to the influence brought by this transparency fluctuation, resulting in a decrease in detection robustness and accuracy. The printed marks on the outer surface of the insulating layer also interfere with defect detection, resulting in misjudging the printed marks as defects.

[0018] For example, assume that an on-line defect detection system for semi-transparent insulated wires based on camera-captured wire images is used on a high-speed production line. Due to the semi-transparency of the insulating layer, not only scratches on the surface of the insulating layer but also faint bubbles inside the insulating layer or textures on the surface of the conductor may be visible in the image. If the transparency of the insulating layer increases due to changes in production batches or process parameters, the normal textures on the surface of the conductor will be more obvious in the image, resulting in the normal textures on the surface of the conductor being misidentified as defects. Conversely, if the transparency decreases, the bubbles inside the insulating layer may become blurred, resulting in the bubbles being missed. At the same time, the model printed marks on the surface of the wire form specific patterns in the image, and these patterns may be similar to the image features of some defects, resulting in the detection system misjudging the printed marks as defects. These problems are particularly prominent in a high-speed production environment and require the detection system to be able to process quickly and accurately.

[0019] If the above problems are not solved, it will lead to a high false alarm rate and missed alarm rate in the defect detection system for semi-transparent insulating conductors. False alarms will increase the subsequent manual re-inspection cost and reduce production efficiency; missed alarms may cause defective products to flow into the market, affecting product quality and user safety, and damaging the enterprise's reputation. This directly affects the application effect of the automated detection system in the production quality control of semi-transparent insulating conductors.

[0020] In response to this, on the first aspect, as Figure 1 shown, the present application proposes a real-time detection method for conductor surface defects for detecting surface defects of semi-transparent insulating conductors, which includes the following steps: S1. Obtain a multi-spectral image sequence of a semi-transparent insulating conductor, where the multi-spectral image sequence includes optical images with different optical wavelengths; S2. Obtain the transparency information of the insulating layer according to the optical characteristics of the optical image with the preset wavelength, and then determine the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer; S3. Use the region recognition algorithm to identify the surface printing marks in the optical images with optical wavelengths within the preset band to obtain the surface printing identification region; S4. Extract features from the regions other than the surface printing identification region in all optical images to obtain a first feature map reflecting different layer information; S5. Use the first defect detection algorithm to identify defects in the first feature map, and then when there are surface defects, determine the location of the surface defect according to the feature differences of the same surface defect in different optical images.

[0021] Among them, the semi-transparent insulating conductor refers to a conductor covered with an insulating layer having a certain light transmittance, and its characteristics enable the simultaneous capture of surface, internal, and conductor layer information during image acquisition, posing challenges for defect detection. The multi-spectral image sequence refers to a series of optical images acquired at different optical wavelengths, and its purpose is to obtain characteristic information reflecting different layers (surface, internal, conductor) of the conductor through the differences in the penetration and reflectivity of materials at different wavelengths. This embodiment can use a multi-spectral camera or a filter set combined with a camera to obtain the multi-spectral image sequence.

[0022] The optical characteristics of an optical image with an optical wavelength of a preset wavelength refer to the optical properties exhibited by the optical image with an optical wavelength of the preset wavelength in a multi-spectral image sequence (equivalent to the optical properties exhibited by an image acquired at a specific selected optical wavelength). Such optical characteristics can be pixel values, color values, texture features, etc. Since the optical characteristics of the optical image are associated with the transparency of the insulating layer, for example, if the insulating layer has a higher transparency, the image will be brighter; conversely, if the transparency is lower, the image will be darker. Therefore, in this embodiment, the transparency information of the insulating layer can be obtained based on the optical characteristics of the optical image with an optical wavelength of the preset wavelength. It should be understood that because the insulating layer of a semi-transparent insulating conductor has different absorption and scattering characteristics for light of different wavelengths, and at a specific wavelength, the absorption coefficient or scattering coefficient of the insulating layer will change significantly with a slight change in transparency, that is, even if there are slight fluctuations in the transparency of the insulating layer, the optical characteristics (such as pixel gray values, brightness, contrast, etc.) of the optical image acquired at this specific wavelength will change correspondingly and significantly. Therefore, the preset wavelength in this embodiment is preferably a wavelength that is more sensitive to changes in the transparency of the insulating layer. Step S2 can obtain the transparency information of the insulating layer by querying a pre-constructed mapping relationship table of optical characteristics and transparency based on the optical characteristics of the optical image with an optical wavelength of the preset wavelength. Step S2 can also obtain the transparency information of the insulating layer by inputting the optical characteristics of the optical image with an optical wavelength of the preset wavelength into a pre-trained transparency evaluation model. This transparency evaluation model is used to evaluate the transparency of the insulating layer based on the input optical characteristics and generate and output the transparency information of the insulating layer. The transparency information of the insulating layer in this embodiment refers to a quantitative description of the light-transmitting degree of the insulating layer. Since step S2 determines the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm based on the transparency information of the insulating layer, that is, step S2 confirms the transparency information of the insulating layer as the parameters of the region recognition algorithm and the first defect detection algorithm, so step S2 is equivalent to adaptively adjusting the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer, so that the region recognition algorithm and the first defect detection algorithm can adapt to the current transparency of the insulating layer, that is, even if there are fluctuations in the transparency of the insulating layer material, this embodiment can accurately identify the surface printed identification area and surface defects. The parameters of the region recognition algorithm and the parameters of the first defect detection algorithm refer to the configuration values or settings used to control the behavior and performance of the region recognition algorithm and the first defect detection algorithm, and their purpose is to enable the algorithm to adaptively adjust its processing method according to the transparency information of the insulating layer to improve the robustness of the detection.

[0023] The surface print mark recognition in step S3 refers to automatically recognizing and locating the print marks existing on the surface of the translucent insulating conductor by using the area recognition algorithm with completed parameter confirmation. The purpose is to exclude these print marks from the defect detection range to avoid misjudging the print marks as defects. Specifically, the area recognition algorithm in this embodiment can adopt template matching, feature point detection, or deep learning methods to implement the surface print mark recognition for the optical image with the optical wavelength within the preset wavelength band. It should be understood that the preset wavelength is a specific wavelength value, and the preset wavelength band is a wavelength range. Therefore, the object of the surface print mark recognition in step S3 is essentially the optical image corresponding to the optical wavelength with the preset wavelength within the preset wavelength band.

[0024] Since the multi-spectral image sequence contains optical images with different optical wavelengths, and the interaction modes of light with different wavelengths with the translucent insulating conductor are different, that is, the optical images with different optical wavelengths can reflect the information of different layers of the conductor. For example, the optical image with a shorter wavelength (such as ultraviolet light) mainly reflects the information on the surface of the insulating layer (such as scratches or contaminants on the surface of the insulating layer), and the optical image with a longer wavelength (such as infrared light, which can penetrate the insulating layer) can reflect the information inside the insulating layer or on the surface of the conductor (such as internal bubbles, impurities, or textures on the surface of the conductor). Therefore, in step S4, the comprehensive feature representation (the first feature map) that can reflect the information of different layers of the conductor can be obtained by extracting the features of these optical images with different wavelengths and fusing the extracted features, so as to realize obtaining the first feature map reflecting the information of different layers by extracting the features of the areas other than the surface print identification area in all the optical images. Step S4 can adopt edge detection, texture analysis, color space conversion, or deep learning feature extraction to implement the feature extraction of the areas other than the surface print identification area in all the optical images.

[0025] The defect identification in step S5 refers to automatically identifying surface defects in the first feature map using the first defect detection algorithm that has completed parameter confirmation. It should be understood that when no surface defects are identified (i.e., there are no surface defects), step S5 ends. Since the manifestations of the same surface defect in optical images corresponding to different optical wavelengths are different (i.e., the features of the same surface defect in optical images corresponding to different optical wavelengths are different, that is, the appearance, contrast, shape, or visibility, etc. of the same surface defect presented in the optical images collected at different optical wavelengths are different), for example, the defects on the surface of the insulating layer are more obvious in the short-wavelength image, and the defects inside the insulating layer and on the surface of the conductor are more obvious in the long-wavelength image. Therefore, when there are surface defects, step S5 determines the location of the surface defect based on the feature differences of the same surface defect in different optical images to achieve hierarchical positioning of surface defects. And because step S5 determines the location of the surface defect based on the feature differences of the same surface defect in different optical images, this embodiment can effectively avoid the situation of misjudgment of the location caused by only using single-image information for surface defect positioning (such as misjudging the surface defect actually located on the surface of the insulating layer as being located inside the insulating layer or on the surface of the conductor), thereby effectively improving the accuracy and reliability of surface defect positioning.

[0026] The core innovation of this application lies in obtaining a multi-spectral image sequence of a semi-transparent insulating conductor, obtaining the transparency information of the insulating layer based on the optical characteristics of the specific-wavelength image, thereby adaptively adjusting the parameters of the region recognition algorithm and the first defect detection algorithm according to the transparency information, and at the same time using the region recognition algorithm to exclude the surface printing identification area, solving the problem of insufficient detection accuracy caused by the transparency fluctuation of the insulating layer and the interference of printing marks, and achieving the effect of improving the detection robustness and accuracy.

[0027] The solution of this application obtains a multi-spectral image sequence of a semi-transparent insulating conductor, providing multi-dimensional information for subsequent analysis. Then, it obtains the transparency information of the insulating layer based on the optical characteristics of the specific optical wavelength image and dynamically adjusts the parameters of the region recognition algorithm and the first defect detection algorithm with this information, so that the detection process can adapt to the transparency fluctuation of the insulating layer. Subsequently, it uses the adjusted region recognition algorithm to identify and exclude the surface printing identification area on the specific-band image, avoiding the interference of printing marks on defect detection. Then, it extracts the features of the non-printing identification areas in all optical images to obtain the first feature map reflecting information of different layers. Finally, it uses the adjusted first defect detection algorithm to identify defects in the first feature map, and when surface defects are detected, it determines their locations according to the feature differences of the same surface defect in different optical images, achieving accurate identification and positioning of surface defects on the semi-transparent insulating conductor.

[0028] In some preferred embodiments, the multi-spectral image sequence can be obtained by equipping with multiple narrow-band filters or using a multi-spectral camera, such as acquiring images in the visible light and near-infrared bands. The preset wavelength can be selected as a specific wavelength sensitive to the transparency change of the insulating layer material, such as a certain near-infrared wavelength. On the optical image at this preset wavelength, this embodiment can characterize the transparency by calculating the average pixel value as an optical feature. This embodiment determines the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm by querying a pre-constructed mapping relationship table. This mapping relationship table can map these average pixel values to the insulating layer transparency level, and then map the insulating layer transparency level to specific parameter sets (such as thresholds, filter parameters, or model weights) of the region recognition algorithm (such as template matching or feature point detection) and the first defect detection algorithm (such as image segmentation or classification). For example, when the optical feature is the average gray value, it can be set that when the gray value is in the range of 0-50, the transparency level is low; when the gray value is in the range of 51-150, the transparency level is medium; when the gray value is in the range of 151-255, the transparency level is high. When the optical feature is a color feature, it can be set that when the color is biased towards red, the transparency level is low; when the color is biased towards green, the transparency level is medium; when the color is biased towards blue, the transparency level is high. Surface print mark recognition can be performed on the visible light band image, and the region recognition algorithm is used to identify the print mark region. Feature extraction can adopt a multi-channel convolutional neural network to jointly extract features from visible light and near-infrared images. This multi-channel convolutional neural network can be an existing multi-channel weighted convolutional neural network (MC-WCNN), multi-channel one-dimensional convolutional network (MC-1DCNN), multi-channel convolutional attention network (MC-CAN), or depthwise separable multi-channel convolutional network (such as MobileNet and ShuffleNet). Defect recognition can analyze the extracted feature map using the first defect detection algorithm. This first defect detection algorithm can be obtained by training a convolutional neural network using a pre-constructed data set containing various defect samples and non-defect samples, and continuously adjusting the network parameters using the backpropagation algorithm during the training process so that it can learn effective features for distinguishing defects and non-defects. This first defect detection algorithm can also be obtained by fine-tuning a pre-trained model (preferably a convolutional neural network model pre-trained on a large data set such as ImageNet, such as ResNet, VGGNet, or EfficientNet) using a pre-constructed data set containing a small number of defect samples and non-defect samples. When a surface defect is detected, the exact position of the surface defect can be determined by comparing the contrast or morphological differences of the surface defect in the visible light and near-infrared images.

[0029] A real-time detection method for surface defects of a conductor provided by the present application can make the region recognition algorithm and the first defect detection algorithm applicable to the current transparency of the insulating layer by determining the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer. Therefore, the present application can effectively solve the problem that the robustness and accuracy of defect detection decrease due to the difficulty of fixed-parameter defect detection in adapting to the influence brought by such transparency fluctuations. And since the present application can eliminate the interference of printing marks on defect detection by first using the region recognition algorithm to identify the surface printing identification area and then detecting defects in the area other than the surface printing identification area in the optical image, the present application can effectively solve the problem that printing marks interfere with defect detection and lead to misjudging printing marks as defects. Moreover, since the present application can realize defect recognition and positioning at different levels by performing multi-spectral feature extraction and defect recognition on the non-printing identification area and determining its position by using the feature differences of the same surface defect in different optical images, the present application can also effectively solve the problem that the accuracy and reliability of defect detection of semi-transparent insulating conductors are insufficient due to the possible superposition or interference of defects at different levels in the image, making it difficult to accurately distinguish. That is, the present application can effectively improve the accuracy, reliability, and robustness of defect detection.

[0030] Due to the possible local differences in the transparency of the insulating layer, if only globally unified parameters are used for region recognition and defect detection, it may lead to a decrease in the accuracy of region recognition and defect detection. For example, in areas with low transparency, the printing marks may be blurred, resulting in failure to recognize the printed identification; in areas with high transparency, the normal texture inside the insulating layer may be misjudged as a defect.

[0031] To solve this technical problem, in some preferred embodiments, step S2 includes: S21. Obtain the pixel values of each pixel point in the optical image with a preset optical wavelength; S22. Use the clustering algorithm to divide the optical image with a preset optical wavelength into multiple transparency analysis regions according to all pixel values; S23. For each transparency analysis region, query the pre-constructed mapping relation table of optical features and transparency according to its optical features in the optical image with a preset optical wavelength to obtain the insulating layer transparency information; S24. For each transparency analysis region, query the pre-constructed mapping relation table of insulating layer transparency, region recognition algorithm parameters, and defect detection parameters according to the insulating layer transparency information to obtain the corresponding region recognition algorithm and the parameters of the first defect detection algorithm; Step S3 includes: S31. For each transparency analysis region, use the corresponding region recognition algorithm to identify surface printing marks in the transparency analysis region to obtain a surface printing identification region; Step S5 includes: S51. For each transparency analysis region, use the corresponding first defect detection algorithm to identify defects in the region of the transparency analysis region except the surface printing identification region, and then when there are surface defects, determine the location of the surface defect according to the characteristic differences of the same surface defect in different optical images.

[0032] Among them, the clustering algorithm refers to an unsupervised learning algorithm used to divide a data set into several clusters according to the similarity between samples, so that the samples within the same cluster have a high degree of similarity, and the samples between different clusters have a low degree of similarity. The clustering algorithm in this embodiment can adopt algorithms such as K-means algorithm, mean shift clustering algorithm or spectral clustering algorithm. This embodiment can divide an optical image with an optical wavelength of a preset wavelength into multiple transparency analysis regions according to all pixel values by using the clustering algorithm, so as to divide the optical image into regions with relatively consistent pixel values according to the optical characteristics reflected by the pixel values. Since the transparency is associated with the pixel values, this embodiment is equivalent to dividing the optical image into regions with relatively consistent transparency. The transparency analysis region refers to an image sub-region obtained by division using the clustering algorithm. The transparency of the insulating layer within each transparency analysis region has a certain degree of consistency. The transparency analysis region is the basic unit for local transparency analysis and parameter adjustment. The mapping relationship table of optical characteristics and transparency in this embodiment refers to a look-up table that stores the relationships between various optical characteristics and the corresponding insulating layer transparency values. The mapping relationship table can be established by collecting images and analyzing characteristics of insulating layer samples with known transparency, and its purpose is to quickly obtain the transparency information of the insulating layer according to the optical characteristics of the optical image. The mapping relationship table of insulating layer transparency, region recognition algorithm parameters and defect detection parameters in this embodiment refers to a look-up table that stores the relationships between different insulating layer transparency values and the parameter sets of the region recognition algorithm and the first defect detection algorithm optimized for this transparency. For example, for a low transparency region, the similarity threshold of the region recognition algorithm (such as template matching based on correlation coefficient) can be set to a lower value, and the edge intensity threshold of the first defect detection algorithm (such as detection based on edge intensity) can be set to a lower value; for a high transparency region, the similarity threshold of the region recognition algorithm can be set to a higher value, the edge intensity threshold of the first defect detection algorithm can be set to a higher value, and the texture filtering parameter can be enabled; for a medium transparency region, the parameter settings are between the two. The mapping relationship table can be established by testing and adjusting algorithm parameters on different transparency samples, and its purpose is to determine the algorithm parameters suitable for this transparency according to the insulating layer transparency information.

[0033] In this embodiment, the adaptive segmentation of the optical image with the preset optical wavelength is realized by using a clustering algorithm to divide the optical image with the preset optical wavelength into multiple transparency analysis regions according to all pixel values, so that the transparency of the insulating layer in each transparency analysis region is relatively consistent, overcoming the challenges brought by the local differences in the transparency of the insulating layer. This embodiment can determine the transparency information of the insulating layer according to its optical characteristics for each transparency analysis region, and then determine the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer, so as to adaptively adjust the parameters of the region recognition algorithm and the first defect detection algorithm according to the local differences in the transparency of the insulating layer. For example, in regions with lower transparency, parameters more sensitive to fuzzy features can be used for print mark recognition and defect detection to reduce missed detections; in regions with higher transparency, parameters insensitive to internal textures can be used to reduce false positives. Therefore, this embodiment can effectively eliminate the influence of the local differences in the transparency of the insulating layer on the region recognition accuracy and defect detection accuracy, thereby effectively improving the accuracy and reliability of print mark recognition and defect detection.

[0034] Since the texture differences on the surface of the insulating layer also affect the judgment of transparency. For example, rough surface areas may scatter light, resulting in an underestimated transparency, while smooth surface areas may overestimate the transparency. This error will further affect the parameter selection of the subsequent region recognition algorithm and the first defect detection algorithm, reducing the accuracy of print mark recognition and defect detection. Therefore, the above embodiment has the problem that the accuracy of the transparency information of the insulating layer is low and the accuracy of print mark recognition and defect detection decreases because the influence of the surface texture of the insulating layer on the transparency of the insulating layer is not considered.

[0035] To solve this technical problem, in some preferred embodiments, step S23 includes: S231. For each transparency analysis region, query the pre-constructed mapping relation table of optical characteristics and transparency according to its optical characteristics in the optical image with the preset optical wavelength to obtain preliminary transparency information; S232. For each transparency analysis region, calculate the actual gray variance used to characterize the surface texture according to the gray values of all pixel points it contains, and then query the pre-constructed mapping relation table of gray variance and transparency compensation amount according to the actual gray variance to obtain the first transparency compensation amount; S233. For each transparency analysis region, calculate the transparency information of the insulating layer according to the corresponding preliminary transparency information and the first transparency compensation amount.

[0036] Among them, the actual gray variance of this embodiment refers to the statistical variance of the gray values of all pixel points within the transparency analysis region, which can be specifically obtained by calculating the average value of the squares of the differences between the gray values of all pixels within the transparency analysis region and their average gray value. Its purpose is to quantify the dispersion degree of the pixel gray values within the region. Since when the surface texture is relatively smooth, the reflection or transmission of light in this transparency analysis region is relatively uniform, and the gray value differences of each pixel point are small, the actual gray variance calculated at this time is also small; when the surface texture is relatively rough, the reflection or transmission of light in this transparency analysis region will be scattered due to the unevenness of the surface, and the gray value differences of each pixel point are large, and the actual gray variance calculated at this time is large. Therefore, the actual gray variance of this embodiment can characterize the surface texture. The mapping relationship table of the gray variance and the transparency compensation amount of this embodiment stores the transparency compensation amounts corresponding to different gray variances. For example, when the gray variance is 0 - 10, the transparency compensation amount is 0; when the gray variance is 10 - 20, the transparency compensation amount is 0.02; when the gray variance is 20 - 30, the transparency compensation amount is 0.04; when the gray variance is 30 - 40, the transparency compensation amount is 0.05... Step S233 can calculate the transparency information of the insulating layer by directly summing or weighted summing the preliminary transparency information corresponding to the transparency analysis region and the first transparency compensation amount. This embodiment is equivalent to not only considering the optical characteristics of the insulating layer material itself when obtaining the transparency information of the insulating layer, but also introducing the quantification and correction of the influence of the surface texture, so as to obtain more accurate transparency information of the insulating layer and provide a more reliable parameter basis for subsequent region recognition and defect detection algorithms. Therefore, this embodiment can effectively solve the problem that the accuracy of the transparency information of the insulating layer is low due to not considering the influence of the surface texture of the insulating layer on the transparency of the insulating layer, and the accuracy of print mark recognition and defect detection decreases, thereby further improving the accuracy of print mark recognition and defect detection.

[0037] In practical applications, the ambient light intensity will also affect the brightness and contrast of the image, and thus affect the accuracy of transparency analysis. If the influence of ambient light is ignored, it may lead to inaccurate transparency evaluation and ultimately affect the reliability of defect detection.

[0038] To solve this technical problem, in some preferred embodiments, step S233 includes: A1. Obtain the ambient light intensity information, and then query the pre - constructed mapping relationship table of the ambient light intensity and the transparency compensation amount according to the ambient light intensity information to obtain the second transparency compensation amount; A2. For each transparency analysis region, calculate the transparency information of the insulating layer according to the second transparency compensation amount, the corresponding preliminary transparency information, and the first transparency compensation amount.

[0039] Among them, the ambient light intensity information refers to a value or signal reflecting the light intensity of the detection environment, which can be obtained by using a light sensor. The mapping relationship table of the ambient light intensity and the transparency compensation amount in this embodiment refers to a data table established through experiments or calibration. This mapping relationship table records the transparency compensation amounts required under different ambient light intensities. For example, in the case of low light intensity, such as 100 Lux, the overall image is dark, and the transparency is underestimated. At this time, the mapping relationship table may suggest a positive transparency compensation amount, such as +0.05, to increase the transparency evaluation value; in the case of moderate light intensity, such as 500 Lux, the image brightness is moderate, and the transparency evaluation is relatively accurate. At this time, the mapping relationship table may suggest a transparency compensation amount close to zero, such as 0 or +0.01, for fine-tuning; in the case of high light intensity, such as 1000 Lux, the overall image is bright, and the transparency is overestimated. At this time, the mapping relationship table may suggest a negative transparency compensation amount, such as -0.03, to reduce the transparency evaluation value. The second transparency compensation amount in this embodiment refers to a correction value used to correct the influence of ambient light on the transparency evaluation. Step A1 can calculate the transparency information of the insulating layer by directly summing or weighted summing the preliminary transparency information, the first transparency compensation amount, and the second transparency compensation amount corresponding to the transparency analysis region. This embodiment is equivalent to making the calculated transparency information of the insulating layer effectively eliminate the influence brought by the ambient light fluctuation by introducing the ambient light intensity information and compensating the preliminary transparency information based on the ambient light intensity information, so as to avoid the situation that the accuracy of the insulating layer transparency information, the printed mark recognition, and the defect detection decreases due to not considering the influence of the ambient light intensity on the insulating layer transparency, that is, this embodiment can further improve the accuracy of the printed mark recognition and the defect detection.

[0040] Since different insulating layer materials have different sensitivities to light of different wavelengths, that is, light of some wavelengths may be more likely to penetrate a certain insulating layer material, while light of other wavelengths cannot penetrate a certain insulating layer material. If a fixed preset wavelength is used, the pixel values of each pixel point in the optical image with the optical wavelength being the preset wavelength cannot accurately reflect the transparency information of different materials, thereby affecting the accuracy of subsequent defect detection.

[0041] To solve this technical problem, in some preferred embodiments, step S21 includes: S211. Obtain the insulating layer material type information, and then query the pre-constructed mapping relationship table of the material type and the transparency change sensitive wavelength according to the insulating layer material type information to obtain the preset wavelength; S212. Obtain the pixel values of each pixel point in the optical image with the optical wavelength being the preset wavelength.

[0042] Among them, the insulating layer material type information refers to the specific types of materials used to manufacture the insulating layer of the semi-transparent insulating conductor, such as polyethylene, polyvinyl chloride, polyurethane, etc. It can be obtained by manual input, reading product identification, or through material identification sensors, etc. The purpose is to distinguish the optical characteristics of different insulating layer materials. The pre-constructed mapping relationship table between the material type and the wavelength sensitive to the transparency change refers to a data table that stores the corresponding relationship between different insulating layer material types and the optical wavelengths that are most sensitive to the transparency change of each. For example, for polyethylene (PE) material, the wavelength sensitive to the transparency change is 650 nm; for polyvinyl chloride (PVC) material, the wavelength sensitive to the transparency change is 520 nm; for polypropylene (PP) material, the wavelength sensitive to the transparency change is 1450 nm. This mapping relationship table can be determined by experimentally measuring the transmittance or absorbance change characteristics of different materials at different wavelengths. The purpose is to select the most suitable optical wavelength for transparency analysis for different insulating layer materials. The preset wavelength in this embodiment refers to the optical wavelength that is most sensitive to the transparency change of the current insulating layer material, queried from the mapping relationship table according to the insulating layer material type information. This preset wavelength is the specific wavelength used to collect the optical images required for subsequent transparency analysis. The purpose is to ensure that the acquired image data can accurately reflect the transparency characteristics of the current insulating layer material. This embodiment is equivalent to dynamically selecting the optical wavelength that is most sensitive to the transparency change of the insulating layer material type to obtain the image pixel values and optical characteristics, that is, this embodiment can more accurately capture and reflect the true transparency characteristics of different insulating layer materials, providing a more reliable data basis for subsequent acquisition of insulating layer transparency information, region division, and adjustment of algorithm parameters based on transparency, thereby significantly improving the adaptability of print mark recognition and defect detection to different insulating layer materials, and further enhancing the accuracy and reliability of defect detection and effectively avoiding misjudgment and missed detection problems caused by material differences.

[0043] In some preferred embodiments, the real-time conductor surface defect detection method further includes steps performed after step S3, including: S6. Query the pre-constructed mapping relationship table between the print mark image features and the defect detection algorithm parameters according to the image features of the surface print mark area to determine the parameters of the second defect detection algorithm; S7. Extract features from the surface print mark areas in all optical images to obtain a second feature map that can reflect information on different levels, and then use the second defect detection algorithm to identify defects in the second feature map.

[0044] Among them, the image features of the surface printed identification area refer to the data used to describe the visual attributes of the surface printed identification area, which can be represented by color distribution, texture features, shape features, or a combination thereof. The pre-constructed mapping relationship table between the printed identification image features and the defect detection algorithm parameters refers to a data table storing the association relationship between different printed identification image features and the corresponding preferred defect detection algorithm parameters. For example, when the printed identification image feature is that the color of the printed mark is dark blue and the texture is relatively simple, the defect detection algorithm parameter is to select the defect detection algorithm based on color threshold segmentation and the color threshold is a value close to dark blue; when the printed identification image feature is that the color of the printed mark is light gray and the texture is relatively complex, the defect detection algorithm parameter is to select the defect detection algorithm based on texture analysis and use the Gabor filter to extract texture features. The parameters of the second defect detection algorithm refer to the settings used to configure the second defect detection algorithm to adapt it to the image features of a specific surface printed identification area, which can include thresholds, filter parameters, model weights, or algorithm flow control parameters. Extracting the features of the surface printed identification area in all optical images refers to extracting the information that can characterize the characteristics of the area from the image data corresponding to the surface printed identification area in the optical images of different optical wavelengths, which can be achieved by methods such as filtering, transformation, statistical analysis, or deep learning. The second feature map of this embodiment refers to the data representation obtained through feature extraction and containing the visual or physical characteristics of the surface printed identification area at different optical wavelengths, which can be a feature vector or a feature image composed of multiple channels or dimensions. The second defect detection algorithm refers to an algorithm specifically used to detect defects in the surface printed identification area, which can be implemented by methods such as rule-based, statistical, machine learning, or deep learning. Defect identification refers to the process of using the second defect detection algorithm to determine whether there are defects in the surface printed identification area based on the second feature map. This embodiment can first determine the parameters of the second defect detection algorithm according to the image features of the surface printed identification area, then extract the features of the surface printed identification area in all optical images to obtain the second feature map that can reflect information at different levels, and finally use the second defect detection algorithm to perform defect identification on the second feature map to achieve defect detection of the surface printed identification area while overcoming the interference of the printed mark to defect detection. Therefore, this embodiment can identify the defects in the surface printed identification area without misjudging the printed identification as a defect, thereby effectively avoiding the situation where the defects in the surface printed identification area are missed due to the lack of defect detection of the surface printed identification area.

[0045] In some preferred embodiments, step S1 includes: S11. Obtain a multi-spectral image sequence of the semi-transparent insulating conductor; S12. Preprocess the multi-spectral image sequence.

[0046] The preprocessing in this embodiment refers to various operations performed on the multi-spectral image sequence after the acquisition of the multi-spectral image sequence, which may specifically include, but are not limited to, image denoising, contrast enhancement, and brightness correction, etc. This embodiment can improve the image quality of the multi-spectral image sequence by preprocessing the multi-spectral image sequence, so as to provide more accurate and reliable input data for subsequent algorithm parameter determination, surface printing mark recognition, feature extraction of different layer information, and the final defect recognition process, thereby effectively avoiding the error accumulation caused by the original image quality problem, and further effectively improving the accuracy and robustness of the real-time conductor surface defect detection method.

[0047] In some preferred embodiments, the preprocessing includes image registration and distortion correction. The image registration in this embodiment refers to the process of aligning multiple images of the same scene obtained at different times, by different sensors, from different perspectives, or in different bands in terms of spatial position through spatial transformation, which can be implemented by using a feature point-based registration method or a region-based registration method. This embodiment can solve the possible misalignment problem between different optical images in the multi-spectral image sequence by aligning the multi-spectral image sequence. The distortion correction in this embodiment refers to the process of eliminating or reducing the geometric distortion of the image caused by factors such as the optical system or sensor during the image acquisition process through a distortion correction model or algorithm. This embodiment can implement the distortion correction of the multi-spectral image sequence by first using the camera calibration method to obtain the camera internal parameters and distortion coefficients, and then applying these parameters to perform distortion recognition and correction on each optical image. This embodiment can compensate for the curvature distortion caused by the change in the wire body diameter or thickness by performing distortion correction on the multi-spectral image sequence, thereby effectively eliminating or suppressing the curvature distortion caused by the change in the wire body diameter or thickness under the influence of factors such as the characteristics of the extrusion process and the flexibility of the conductor itself, and the situation where the printing mark recognition and defect detection are affected.

[0048] As can be seen from the above, a real-time detection method for conductor surface defects provided by this application can make the region recognition algorithm and the first defect detection algorithm applicable to the current transparency of the insulating layer by determining the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer. Therefore, this application can effectively solve the problem that the robustness and accuracy of defect detection decrease due to the difficulty of fixed-parameter defect detection in adapting to the influence brought by such transparency fluctuations. Since this application can eliminate the interference of printing marks on defect detection by first using the region recognition algorithm to identify the surface printing identification region and then detecting defects in the region other than the surface printing identification region in the optical image, this application can effectively solve the problem that printing marks interfere with defect detection and lead to misjudging printing marks as defects. And since this application can realize defect recognition and positioning at different levels by extracting multi-spectral features and defect recognition for non-printing identification regions and using the feature differences of the same surface defect in different optical images to determine its position, this application can also effectively solve the problem that the accuracy and reliability of defect detection of semi-transparent insulating conductors are insufficient due to the possible superposition or interference of defects at different levels in the image, making it difficult to accurately distinguish. That is, this application can effectively improve the accuracy, reliability, and robustness of defect detection.

[0049] In a second aspect, as Figure 2 shown, this application also provides a real-time detection system for conductor surface defects, which is used for detecting surface defects of a semi-transparent insulating conductor and includes: An image sequence acquisition module 1, which is used to acquire a multi-spectral image sequence of the semi-transparent insulating conductor, and the multi-spectral image sequence includes optical images with different optical wavelengths; An algorithm parameter confirmation module 2, which is used to obtain the transparency information of the insulating layer according to the optical characteristics of the optical image with a preset wavelength, and then determine the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer; A printing identification region recognition module 3, which is used to use the region recognition algorithm to identify surface printing marks in the optical image with an optical wavelength within a preset band to obtain the surface printing identification region; A feature extraction module 4, which is used to extract features from the regions other than the surface printing identification region in all optical images to obtain a first feature map reflecting information at different levels; A defect detection module 5, which is used to use the first defect detection algorithm to identify defects in the first feature map, and then when there are surface defects, determine the position where the surface defect is located according to the feature differences of the same surface defect in different optical images.

[0050] A real-time conductor surface defect detection system provided by this application includes an image sequence acquisition module 1, an algorithm parameter confirmation module 2, a printed identification area recognition module 3, a feature extraction module 4, and a defect detection module 5. The real-time conductor surface defect detection system provided by this embodiment is used to execute the steps in the real-time conductor surface defect detection method provided in the first aspect above. The principle of the real-time conductor surface defect detection system provided by this embodiment is the same as that of the real-time conductor surface defect detection method provided in the first aspect above, and will not be elaborated in detail here.

[0051] In some preferred embodiments, the process of obtaining the insulation layer transparency information based on the optical characteristics of the optical image with the optical wavelength being the preset wavelength, and then determining the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm includes: B1. Obtain the pixel values of each pixel point in the optical image with the optical wavelength being the preset wavelength; B2. Use the clustering algorithm to divide the optical image with the optical wavelength being the preset wavelength into multiple transparency analysis regions according to all pixel values; B3. For each transparency analysis region, query the pre-constructed mapping relation table of optical characteristics and transparency according to its optical characteristics in the optical image with the optical wavelength being the preset wavelength to obtain the insulation layer transparency information; B4. For each transparency analysis region, query the pre-constructed mapping relation table of insulation layer transparency, region recognition algorithm parameters, and defect detection parameters according to the insulation layer transparency information to obtain the corresponding region recognition algorithm and the parameters of the first defect detection algorithm; The process of using the region recognition algorithm to identify the surface printing mark on the optical image with the optical wavelength in the preset band to obtain the surface printed identification area includes: For each transparency analysis region, use the corresponding region recognition algorithm to identify the surface printing mark on this transparency analysis region to obtain the surface printed identification area; The process of using the first defect detection algorithm to identify defects in the first feature map, and then when there are surface defects, determining the position of the surface defect according to the feature differences of the same surface defect in different optical images includes: For each transparency analysis region, use the corresponding first defect detection algorithm to identify defects in the region of this transparency analysis region except the surface printed identification area, and then when there are surface defects, determine the position of the surface defect according to the feature differences of the same surface defect in different optical images.

[0052] As can be seen from the above, a real-time conductor surface defect detection method and system provided by the present application can make the region recognition algorithm and the first defect detection algorithm applicable to the current transparency of the insulating layer by determining the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer. Therefore, the present application can effectively solve the problem that the robustness and accuracy of defect detection decrease due to the difficulty of fixed-parameter defect detection in adapting to the influence brought by such transparency fluctuations. Since the present application can eliminate the interference of printing marks on defect detection by first using the region recognition algorithm to identify the surface printing identification area and then detecting defects in the area other than the surface printing identification area in the optical image, the present application can effectively solve the problem that printing marks interfere with defect detection and lead to misjudging printing marks as defects. Moreover, since the present application can realize defect recognition and positioning at different levels by extracting multi-spectral features and defect recognition for the non-printing identification area and determining its position by using the feature differences of the same surface defect in different optical images, the present application can also effectively solve the problem that the accuracy and reliability of defect detection of semi-transparent insulating conductors are insufficient due to the possible superposition or interference of defects at different levels in the image and the difficulty in accurately distinguishing them. That is, the present application can effectively improve the accuracy, reliability, and robustness of defect detection.

[0053] In the embodiments provided by the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only illustrative. For example, the above division of units is only a logical function division, and there can be other division methods in actual implementation. For another example, multiple units or components can be combined or integrated into another robot, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections between each other can be through some communication interfaces. The indirect couplings or communication connections of devices or units can be in electrical, mechanical, or other forms.

[0054] In addition, in each embodiment of the present application, the functional modules can be integrated together to form an independent part, or each module can exist alone, or two or more modules can be integrated to form an independent part.

[0055] In this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0056] The above are only the embodiments of the present application and are not intended to limit the protection scope of the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A real-time detection method for surface defects of a conductor, characterized in that, For surface defect detection of a translucent insulating conductor, the real-time detection method for surface defects of the conductor includes the following steps: S1. Obtain a multi-spectral image sequence of the translucent insulating conductor, where the multi-spectral image sequence includes optical images with different optical wavelengths; S2. Obtain the transparency information of the insulating layer according to the optical characteristics of the optical image with the optical wavelength being a preset wavelength, and then determine the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer; S3. Use the region recognition algorithm to identify surface printing marks in the optical images with the optical wavelength within a preset band to obtain the surface printing identification region; S4. Extract features from the regions in all the optical images except the surface printing identification region to obtain a first feature map reflecting information of different levels; S5. Use the first defect detection algorithm to identify defects in the first feature map, and then when there are surface defects, determine the position where the surface defect is located according to the feature differences of the same surface defect in different optical images.

2. The real-time detection method for conductor surface defects according to claim 1, wherein, Step S2 includes: S21. Obtain the pixel values of each pixel point in the optical image with the optical wavelength being a preset wavelength; S22. Use a clustering algorithm to divide the optical image with the optical wavelength being a preset wavelength into multiple transparency analysis regions according to all the pixel values; S23. For each transparency analysis region, query a pre-constructed mapping relationship table of optical characteristics and transparency according to its optical characteristics in the optical image with the optical wavelength being a preset wavelength to obtain the transparency information of the insulating layer; S24. For each transparency analysis region, query a pre-constructed mapping relationship table of insulating layer transparency, region recognition algorithm parameters, and defect detection parameters according to the transparency information of the insulating layer to obtain the corresponding region recognition algorithm and the parameters of the first defect detection algorithm; Step S3 includes: S31. For each transparency analysis region, use the corresponding region recognition algorithm to identify surface printing marks in the transparency analysis region to obtain the surface printing identification region; Step S5 includes: S51. For each transparency analysis region, use the corresponding first defect detection algorithm to identify defects in the region of the transparency analysis region except the surface printing identification region, and then when there are surface defects, determine the position where the surface defect is located according to the feature differences of the same surface defect in different optical images.

3. The real-time detection method for conductor surface defects according to claim 2, wherein, Step S23 includes: S231. For each transparency analysis region, query a pre-constructed mapping relationship table of optical characteristics and transparency according to its optical characteristics in the optical image with the optical wavelength being a preset wavelength to obtain preliminary transparency information; S232. For each transparency analysis region, calculate the actual gray variance used to characterize the surface texture according to the gray values of all the pixel points it contains, and then query a pre-constructed mapping relationship table of gray variance and transparency compensation amount according to the actual gray variance to obtain the first transparency compensation amount; S233. For each of the transparency analysis regions, calculate the transparency information of the insulating layer based on the corresponding preliminary transparency information and the first transparency compensation amount.

4. The real-time detection method for conductor surface defects according to claim 3, wherein Step S233 includes: A1. Obtain the ambient light intensity information, and then query a pre-constructed mapping relationship table of ambient light intensity and transparency compensation amount according to the ambient light intensity information to obtain the second transparency compensation amount; A2. For each of the transparency analysis regions, calculate the transparency information of the insulating layer based on the second transparency compensation amount, the corresponding preliminary transparency information, and the first transparency compensation amount.

5. The real-time detection method for conductor surface defects according to claim 2, characterized in that Step S21 includes: S211. Obtain the insulating layer material type information, and then query a pre-constructed mapping relationship table of material type and transparency change sensitive wavelength according to the insulating layer material type information to obtain the preset wavelength; S212. Obtain the pixel values of each pixel point in the optical image with the optical wavelength being the preset wavelength.

6. The real-time detection method for conductor surface defects according to claim 1, wherein The real-time conductor surface defect detection method further includes steps executed after step S3: S6. Query a pre-constructed mapping relationship table of printing identification image features and defect detection algorithm parameters according to the image features of the surface printing identification region to determine the parameters of the second defect detection algorithm; S7. Extract features from the surface printing identification regions in all the optical images to obtain a second feature map that can reflect information of different layers, and then use the second defect detection algorithm to identify defects in the second feature map.

7. The real-time detection method for conductor surface defects according to claim 1, wherein Step S1 includes: S11. Obtain the multi-spectral image sequence of the semi-transparent insulating conductor; S12. Perform preprocessing on the multi-spectral image sequence.

8. The real-time detection method for conductor surface defects according to claim 7, wherein The preprocessing includes image registration and distortion correction.

9. A real-time detection system for surface defects of a conductor, characterized in that, For surface defect detection of a semi-transparent insulating conductor, the real-time conductor surface defect detection system includes: An image sequence acquisition module, configured to obtain the multi-spectral image sequence of the semi-transparent insulating conductor, where the multi-spectral image sequence includes optical images with different optical wavelengths; An algorithm parameter confirmation module, configured to obtain the transparency information of the insulating layer according to the optical characteristics of the optical image with the optical wavelength being the preset wavelength, and then determine the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm according to the transparency information of the insulating layer; A printing identification region recognition module, configured to use the region recognition algorithm to perform surface printing mark recognition on the optical images with the optical wavelength within the preset band to obtain the surface printing identification region; A feature extraction module, configured to extract features from the regions in all the optical images except the surface printing identification regions to obtain a first feature map that reflects information of different layers; A defect detection module, configured to use the first defect detection algorithm to identify defects in the first feature map, and then determine the position of the surface defect according to the feature differences of the same surface defect in different optical images when there is a surface defect.

10. The real-time conductor surface defect detection system according to claim 9, characterized in that, The process of obtaining the transparency information of the insulating layer according to the optical characteristics of the optical image with the optical wavelength being the preset wavelength, and then determining the parameters of the region recognition algorithm and the parameters of the first defect detection algorithm includes: B1. Obtain the pixel values of each pixel point in the optical image with the optical wavelength being the preset wavelength; B2. Use a clustering algorithm to divide the optical image with the optical wavelength being the preset wavelength into multiple transparency analysis regions according to all the pixel values; B3. For each transparency analysis region, query the pre-constructed mapping relationship table of optical characteristics and transparency according to its optical characteristics in the optical image with the optical wavelength being the preset wavelength to obtain the transparency information of the insulating layer; B4. For each transparency analysis region, query the pre-constructed mapping relationship table of insulating layer transparency, region recognition algorithm parameters, and defect detection parameters according to the transparency information of the insulating layer to obtain the corresponding region recognition algorithm and the parameters of the first defect detection algorithm; The process of using the region recognition algorithm to identify the surface printing marks in the optical image with the optical wavelength in the preset band to obtain the surface printing identification region includes: For each transparency analysis region, use the corresponding region recognition algorithm to identify the surface printing marks in this transparency analysis region to obtain the surface printing identification region; The process of using the first defect detection algorithm to identify defects in the first feature map, and then when there are surface defects, determining the position of the surface defect according to the feature differences of the same surface defect in different optical images includes: For each transparency analysis region, use the corresponding first defect detection algorithm to identify defects in this transparency analysis region, and then when there are surface defects, determine the position of the surface defect according to the feature differences of the same surface defect in different optical images.

Images