Steel Coil Defect Identification Method and Device

Abstract

The present application provides a method, device and electronic device for identifying steel coil defects. The method includes: performing semantic segmentation on a target image through a steel coil defect mask model to obtain an identification result of the steel coil defect area, where the target image is an image including at least part of the end face of the steel coil; based on the identification result of the steel coil defect area, extracting the contours of the steel coil defect areas in the target image to obtain the contours of at least one steel coil defect; calculating the bounding box of each steel coil defect contour to obtain the bounding box size of each steel coil defect; determining whether to generate a defect alarm signal according to the comparison result between the maximum bounding box size among the bounding box sizes of at least one steel coil defect and a preset alarm size; and superimposing the contours of at least one steel coil defect onto the target image. In this way, users can timely and accurately confirm the steel coil defects that seriously affect the quality of the steel coil and the locations of the steel coil defects on the end face of the steel coil.

Description

Technical Field

[0001] The present application relates to the field of image processing technologies, and in particular, to a method, a device, an electronic device, and a computer-readable storage medium for identifying steel coil defects. Background Art

[0002] Edge crack defects and edge loss defects of steel coils refer to small wedge-shaped cracks or even continuous serrated cracks observed on the cross-section after coiling, which are caused by factors such as poor roll adjustment and mismatched roll profiles during strip rolling, resulting in local stress at the edge of the sheet exceeding the material strength limit. Edge crack defects and edge loss defects directly affect the quality of the rolled products leaving the factory. It is very important to detect edge loss defects and edge crack defects in a timely manner for the quality control of steel coils leaving the factory. However, some steel coil defects are within the allowable range and have little impact on the quality of the steel coil, while some steel coil defects exceed the allowable range and seriously affect the quality of the steel coil. Moreover, since edge crack defects and edge loss defects are mostly extremely small openings, they are difficult to observe with the naked eye, and it is impossible to detect in a timely manner whether there are steel coil defects that seriously affect the quality of the steel coil. Therefore, how to detect in a timely manner the steel coil defects that seriously affect the quality of the steel coil has become an urgent task. Summary of the Invention

[0003] The present application provides a method, a device, an electronic device, and a computer-readable storage medium for identifying steel coil defects. Based on the defect alarm signal and the superimposed steel coil end face image, a user can timely and accurately confirm the steel coil defects that seriously affect the quality of the steel coil and the positions where the steel coil defects are located on the steel coil end face.

[0004] According to a first aspect of the present application, there is provided a method for identifying steel coil defects. The method for identifying steel coil defects includes: performing semantic segmentation on a target image through a steel coil defect mask model to obtain an identification result of a steel coil defect area, where the target image is an image including at least a part of the steel coil end face; extracting the contour of the steel coil defect area in the target image based on the identification result of the steel coil defect area to obtain the contour of at least one steel coil defect; calculating the bounding box size of each steel coil defect for each contour of the steel coil defect to obtain the bounding box size of each steel coil defect; determining whether to generate a defect alarm signal according to a comparison result between the maximum bounding box size among the bounding box sizes of at least one steel coil defect and a preset alarm size; and superimposing the contour of at least one steel coil defect onto the target image.

[0005] According to a second aspect of the present application, a steel coil defect recognition device is provided. The steel coil defect recognition device may include: a semantic segmentation module, configured to perform semantic segmentation on a target image through a steel coil defect mask model to obtain an identification result of a steel coil defect area, where the target image is an image including at least a part of the end face of the steel coil; a contour extraction module, configured to extract the contour of the steel coil defect area in the target image based on the identification result of the steel coil defect area to obtain the contour of at least one steel coil defect; a calculation module, configured to perform bounding box calculation on the contour of each steel coil defect to obtain the bounding box size of each steel coil defect; an alarm signal generation determination module, configured to determine whether to generate a defect alarm signal according to the comparison result between the maximum bounding box size among the bounding box sizes of at least one steel coil defect and a preset alarm size; a contour superposition module, configured to superpose the contour of the at least one steel coil defect onto the target image.

[0006] According to a third aspect of the present application, an electronic device is provided. The electronic device includes: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the steps of any one of the above steel coil defect recognition methods are implemented.

[0007] According to a fourth aspect of the present application, a computer-readable storage medium is provided, characterized in that a steel coil defect recognition program instruction is stored on the computer-readable storage medium, and when the steel coil defect alarm program instruction is executed by a processor, any one of the above steel coil defect recognition methods is implemented.

[0008] In summary, the steel coil defect recognition method, device, electronic device, and computer-readable storage medium provided by this application at least have the following beneficial effects: By performing semantic segmentation on the target image through the steel coil defect mask model, the recognition result of the steel coil defect area is obtained. Thus, by capturing the local features of the steel coil end face, the detailed features of the steel coil end face can be expanded, and the steel coil defect mask model can be used to efficiently and accurately identify the defect area in the steel coil image. Based on the recognition result of the steel coil defect area, the contour of the steel coil defect area in the target image is extracted to obtain the contour of at least one steel coil defect. In this way, the precise information of the steel coil defect can be obtained by extracting the contour of the steel coil defect area. By performing bounding box calculation on the contour of each steel coil defect, the position and size of the steel coil defect in the target image can be accurately determined, thereby further precisely measuring the size of the defect. In this way, the shape and size of the defect can be effectively described. And using the bounding box size as the basis for measuring the size of the defect enables defects of different shapes and directions to be described and compared with a unified size parameter, which is conducive to establishing a unified defect alarm standard and making the alarm result more objective. By comparing the maximum bounding box size with the preset alarm size, it is possible to more accurately determine whether a defect alarm signal is generated, which helps to reduce false alarms, improve the reliability and effectiveness of the alarm signal, as well as improve the alarm efficiency and sensitivity. At the same time, the defect alarm signal can promptly and effectively prompt the user of the steel coil defect that seriously affects the quality of the steel coil. Overlaying the contour of at least one steel coil defect onto the target image can help to more intuitively display the specific position and shape of the defect. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the specific embodiments of this application or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings in the following description are some embodiments of this application. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0010] Figure 1 Shows the schematic architecture diagram of a steel coil recognition system provided by an embodiment of this application;

[0011] Figure 2 Shows the schematic flowchart of a steel coil defect recognition method provided by an embodiment of this application;

[0012] Figure 3 and Figure 4 Shows the schematic diagram of an overlaid image provided by an embodiment of this application;

[0013] Figure 5 Shows the schematic structural diagram of a steel coil defect mask model provided by an embodiment of this application;

[0014] Figure 6 and Figure 7 respectively show schematic diagrams of a mask image provided by an embodiment of the present application;

[0015] Figure 8 show a schematic diagram of another superimposed image provided by an embodiment of the present application;

[0016] Figure 9 show a schematic diagram of a target image with a superimposed contour and bounding box provided by an embodiment of the present application;

[0017] Figure 10 show a schematic diagram of another target image with a superimposed contour and bounding box provided by an embodiment of the present application;

[0018] Figure 11 show a schematic flow diagram of another steel coil defect recognition method provided by an embodiment of the present application;

[0019] Figure 12 show a schematic structural diagram of a steel coil defect recognition device provided by an embodiment of the present application; and

[0020] Figure 13 show a schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0021] In order to make the above and other features and advantages of the present application clearer, the present application will be further described below with reference to the accompanying drawings. It should be understood that the specific embodiments given herein are for the purpose of explaining to those skilled in the art and are merely exemplary, not restrictive.

[0022] In the following description, many specific details are set forth to provide a thorough understanding of the present application. However, it is obvious to those skilled in the art that the present application does not need to be practiced with these specific details. In other cases, well-known steps or operations are not described in detail to avoid obscuring the present application.

[0023] On the one hand, an embodiment of the present application provides a steel coil defect recognition system, Figure 1 show a schematic architecture diagram of a steel coil recognition system provided by an embodiment of the present application. As Figure 1 shown, the steel coil defect recognition system 10 may include an image acquisition device 11 and a steel coil defect recognition device 12. Among them, the image acquisition device 11 may include at least two cameras. The multiple cameras may be distributed on both sides of the steel coil and can acquire partial images of different side end faces of the steel coil. The cameras may be high-power zoom high-definition cameras, such as 40x zoom high-definition dome cameras. The steel coil defect recognition device 12 may be used to execute the steps of any one of the steel coil defect recognition methods provided by the embodiments of the present application.

[0024] In one aspect of the embodiments of the present application, a method for identifying steel coil defects is provided. This method for identifying steel coil defects is applied to a steel coil defect identification device 12 as shown in Figure 1 . Figure 2 The flowchart of a method for identifying steel coil defects provided by the embodiments of the present application is shown. As shown in Figure 2 , this method for identifying steel coil defects may include the following steps.

[0025] S21, perform semantic segmentation on the target image through a steel coil defect mask model to obtain the recognition result of the steel coil defect area.

[0026] In an embodiment of the present application, the target image is an image including at least part of the end face of the steel coil, which can be a high-definition image and can clearly show the local details of the end face of the steel coil. The target image can be obtained from the image acquisition device 11. Specifically, the image acquisition device 11 can acquire the initial image of the local part of the end face of the steel coil and send it to the steel coil defect identification device 12. The steel coil defect identification device 12 can perform image enhancement processing on the initial image to obtain a target image with clear image and obvious features.

[0027] The target image can also be obtained from the cloud. Specifically, the image acquisition device 11 uploads the acquired initial image of the local part of the end face of the steel coil to the cloud, and the cloud sends the initial image to the steel coil defect identification device 12. The steel coil defect identification device 12 can perform image enhancement processing on the initial image to obtain a target image with clear image and obvious features.

[0028] In an embodiment of the present application, the "steel coil defect mask model" is a model that has been trained with a set of local images of steel coils labeled with various defects and can perform semantic segmentation and mask information extraction. The "steel coil defect mask model" can be constructed based on a deep learning model. The recognition result of the steel coil defect area can be obtained after the steel coil defect mask model processes the target image. The recognition result of the steel coil defect area may include, but is not limited to, the position, type, defect area, mask information and size, and visualization result of the steel coil defect area.

[0029] In the embodiments of the present application, inputting the target image into the steel coil defect mask model for semantic segmentation can segment the target image into a steel coil defect area and a non-defect area, thereby obtaining the recognition result of the steel coil defect area.

[0030] In another embodiment, the steel coil defect mask model can also identify the type corresponding to the steel coil defect according to the shape of the steel coil defect and label the corresponding defect identifier. In addition, the steel coil defect mask model uses different identifiers to mark the steel coil defect area and the non-defect area. For example, the steel coil defect area is marked as 0, and the non-defect pixels are marked as 1.

[0031] S22. Based on the recognition result of the defective area of the steel coil, extract the contour of at least one defective area of the steel coil to obtain the contour of at least one steel coil defect.

[0032] In one embodiment, the contour extraction can be implemented based on existing contour detection algorithms. Among them, the contour detection algorithm can include an edge detection algorithm.

[0033] In the embodiment of the present application, each edge pixel of the defective area of the steel coil in the recognition result of the defective area of the steel coil is identified through the contour detection algorithm, and the contour of each defective area of the steel coil is obtained according to a set of edge pixels corresponding to each defective area of the steel coil respectively.

[0034] It should be noted that each contour is composed of a series of pixels. Multiple contours can be stored in a contour list for convenient subsequent query of the contour.

[0035] S23. Calculate the bounding box of each contour of the steel coil defect to obtain the bounding box size of each steel coil defect.

[0036] The "bounding box calculation" involved in the embodiment of the present application can include two steps: calculating the bounding box of the enclosing contour and extracting the size of the bounding box. Among them, the bounding box can be the circumscribed box of the contour.

[0037] In one embodiment, the bounding box is a rectangular box, and the size of the bounding box can include the width and height of the bounding box. The implementation method of calculating the bounding box of the enclosing contour is: calculate the convex hull of the contour through the set of pixel points corresponding to the contour. Among them, the convex hull is a polygon that completely contains the original contour and is only composed of the points on the contour. Then, solve the rectangle with the smallest area that can completely enclose the convex hull, and this rectangle is the bounding box. The "bounding box size of the steel coil defect" involved in the embodiment of the present application can correspond to the size of the contour of the steel coil defect.

[0038] In another embodiment, the bounding box is the circumscribed box of other shapes, such as an elliptical box, a trapezoidal box, a triangular box, etc.

[0039] S24. Determine whether to generate a defect alarm signal according to the comparison result between the maximum bounding box size in the bounding box sizes of the at least one steel coil defect and the preset alarm size.

[0040] The "maximum bounding box size" involved in the embodiment of the present application is the bounding box size for determining whether to generate a warning signal among the bounding box sizes of multiple steel coil defects. It should be noted that the maximum bounding box size is the height or the width.

[0041] The preset alarm size is a threshold value representing the maximum tolerable size of defects by the user. Its specific value can be set according to product quality standards and production requirements. In one embodiment, different preset alarm sizes can be set according to the types of candidate defects of the steel coil. The types of candidate defects of the steel coil can include, but are not limited to, edge damage candidate defects, edge crack candidate defects, etc.

[0042] The defect alarm signal can be used to remind the user that there are steel coil defects in a local area of the end face of the steel coil. The missing item alarm signal can include an image alarm signal, a sound alarm signal, an information alarm signal, etc. Among them, the information alarm signal can be realized by displaying an alarm message on the user interface, sending an email, or sending a text message notification. One implementation manner of the image alarm signal is to display a "bad" mark on the superimposed image, such as Figure 3 and 4 shown. Another implementation manner of the image alarm signal is that the steel coil defect area in the superimposed image flashes. Another implementation manner of the image alarm signal is that the steel coil defect area in the superimposed image is marked with a highlighted color.

[0043] In an embodiment of the present application, the steel coil defect recognition device 12 can traverse the bounding box sizes of all steel coil defects, find out the maximum bounding box size, and then compare the size of the maximum bounding box with the alarm size, so as to determine whether to generate a defect alarm signal according to the comparison result.

[0044] S25, superimpose the contours of at least one steel coil defect onto the target image.

[0045] In one embodiment, through an image synthesis algorithm, the contour of each steel coil defect extracted in S22 can be superimposed on the target image, so as to ensure that the contour of each steel coil defect is perfectly fused with the target image.

[0046] Figure 3 and Figure 4 show a schematic diagram of a superimposed image provided by an embodiment of the present application, such as Figure 3 and Figure 4 shown. In the superimposed image, the contour of each steel coil defect can be marked with an obvious line, so that each steel coil defect is visually represented in the superimposed image. In addition, the size of the steel coil defect is marked on the superimposed image for the user to understand the size of each steel coil defect.

[0047] It should be noted that the contours of different types of steel coil defects can be represented by the same line style or different line styles.

[0048] In the above embodiments, semantic segmentation is performed on the target image through the steel coil defect mask model to obtain the recognition result of the steel coil defect area. Thus, by capturing the local features of the steel coil end face, the detailed features of the steel coil end face can be expanded, and the steel coil defect mask model can be used to efficiently and accurately identify the defect area in the steel coil image. Based on the recognition result of the steel coil defect area, the contour of the steel coil defect area in the target image is extracted to obtain the contour of at least one steel coil defect. In this way, accurate information about the steel coil defect can be obtained by extracting the contour of the steel coil defect area. By performing bounding box calculation on the contour of each steel coil defect, the position and size of the steel coil defect in the target image can be accurately determined, thereby further accurately measuring the size of the defect. In this way, the shape and size of the defect can be effectively described. And using the bounding box size as the basis for measuring the size of the defect enables defects of different shapes and directions to be described and compared with a unified size parameter, which is conducive to establishing a unified defect alarm standard and making the alarm result more objective. By comparing the maximum bounding box size with the preset alarm size, it is possible to more accurately determine whether a defect alarm signal is generated, which helps to reduce false alarms, improve the reliability and effectiveness of the alarm signal, as well as improve the alarm efficiency and sensitivity. At the same time, the defect alarm signal can effectively prompt the user in a timely manner that there are steel coil defects that seriously affect the quality of the steel coil. Overlaying the contour of at least one steel coil defect onto the target image can help to more intuitively display the specific position and shape of the defect, and facilitate the user to timely and accurately confirm the steel coil defects existing in the steel coil end face area.

[0049] In some embodiments, S24 can be specifically executed as generating a defect alarm signal when the maximum bounding box size among the bounding box sizes of at least one steel coil defect is greater than the alarm size. When the maximum bounding box size among the bounding box sizes of at least one steel coil defect is not greater than the alarm size, it is determined that no defect alarm signal is generated.

[0050] Specifically, the comparison result can include that the maximum bounding box size is greater than the alarm size and the maximum bounding box size is not greater than the alarm size. The maximum bounding box size being greater than the alarm size may mean that the steel coil defect corresponding to the maximum bounding box size exceeds the range acceptable to the user and belongs to a type of steel coil defect that exceeds the standard, and an alarm needs to be sent to the user. The maximum bounding box size not being greater than the alarm size may mean that the steel coil defects in the local area of the steel coil end face corresponding to the target image are all within the allowable range, and no defect alarm signal needs to be generated.

[0051] In the embodiment of the present application, the steel coil defect recognition device 12 traverses the bounding box sizes of each steel coil defect, compares the bounding box size of each steel coil defect with the alarm size, and determines to generate a defect alarm signal when the bounding box size of a steel coil defect is greater than the alarm size. When the bounding box size of a steel coil defect is not greater than the alarm size, it is determined that no defect alarm signal is generated.

[0052] In the above embodiment, by comparing the maximum bounding box size with the alarm size, it is possible to accurately determine whether there is a defect exceeding the alarm size. A defect alarm signal is generated only when the maximum bounding box size exceeds the alarm size, and a defect alarm signal is generated only when the maximum bounding box size exceeds the alarm size, thereby reducing false alarms or missed alarms and improving the accuracy and reliability of detection.

[0053] The inventor found through a large amount of steel coil quality inspection data that the opening shapes of edge crack defects and edge damage defects are different. The length of edge crack defects exceeding the standard is generally in the range of 5 mm to 100 mm, and the length of edge damage defects exceeding the standard is generally in the range of 10 mm to 200 mm. Therefore, the alarm edge damage size and the alarm edge damage size can be set respectively according to the above characteristic data.

[0054] In some embodiments, S24, determining whether to generate a defect alarm signal according to the comparison result between the maximum bounding box size in the bounding box sizes of the at least one steel coil defect and the preset alarm size includes: when the at least one steel coil defect includes at least one edge damage defect, determining whether to generate an edge damage alarm signal according to the comparison result between the maximum bounding box size in the bounding box sizes of the at least one edge damage defect and the preset alarm edge damage size.

[0055] The defect alarm signal can include alarm signals for various steel coil defects. In one embodiment, the defect alarm signal can include an edge damage alarm signal. The edge damage alarm signal is used to prompt that there is an edge damage defect exceeding the standard in a local area of the steel coil end face. Among them, the edge damage defect can refer to the damaged part or irregular part of the steel coil edge, that is, the damaged part within each layer of the steel coil. The preset alarm edge damage size is a threshold for judging whether an edge damage candidate defect exceeds the standard. In one embodiment of the application, the preset alarm edge damage size is 5 millimeters.

[0056] In the embodiment of the present application, the steel coil defect recognition device 12 first traverses the bounding box sizes of each edge damage candidate defect, selects the maximum bounding box size, and then compares the size of the maximum bounding box with the warning alarm edge damage size. Finally, according to the comparison result, it is determined whether the edge damage defect corresponding to the maximum bounding box size exceeds the standard, so as to determine whether to generate an edge damage alarm signal.

[0057] In the above embodiments, by comparing the maximum bounding box size of the edge damage candidate defect with the alarm edge damage size, it is possible to accurately and timely detect whether there is an edge damage defect beyond the standard range, which is convenient for timely reminding the user that there is a defect beyond the standard in a local area of the steel coil, providing decision-making support for subsequent quality control.

[0058] In some embodiments, S24, according to the comparison result between the maximum bounding box size among the bounding box sizes of the at least one steel coil defect and the preset alarm size, determining whether to generate a defect alarm signal, includes: in the case that the at least one steel coil defect includes at least one edge crack defect, according to the comparison result between the maximum bounding box size among the bounding box sizes of the at least one edge crack defect and the preset alarm edge crack size, determining whether to generate an edge crack alarm signal.

[0059] In one embodiment, the defect alarm signal includes an edge crack alarm signal. The edge crack alarm signal is used to prompt that there is an edge crack defect beyond the standard in a local area of the end face of the steel coil. Among them, the edge crack defect may refer to a crack or fissure that appears at the edge of the steel coil, that is, the open part within each layer of the steel coil. The alarm edge crack size is a threshold for judging whether the edge crack candidate defect exceeds the standard. In an embodiment of the present application, the alarm edge crack size may be 10 millimeters.

[0060] In the embodiment of the present application, the steel coil defect recognition device 12 first traverses the bounding box sizes of each edge crack candidate defect, selects the maximum bounding box size, and then compares the maximum bounding box size with the warning alarm edge damage size. Finally, according to the comparison result, it is determined whether the edge crack defect corresponding to the maximum bounding box size exceeds the standard, so as to determine whether to generate an edge damage alarm signal.

[0061] In the above embodiments, by comparing the maximum bounding box size of the edge damage candidate defect with the alarm edge damage size, it is possible to accurately and timely detect whether there is an edge crack defect beyond the standard range, which is convenient for timely reminding the user that there is a defect beyond the standard in a local area of the steel coil, providing decision-making support for subsequent quality control.

[0062] In order to accurately distinguish the defective part of the steel coil in the target image from the background part, the embodiment of the present application provides a steel coil defect mask model based on a semantic segmentation layer. Figure 5 The structural schematic diagram of a steel coil defect mask model provided by the embodiment of the present application is shown as Figure 5As shown, the steel coil defect mask model 50 may include a semantic segmentation layer 51 and a mask generation layer 52. Among them, the semantic segmentation layer 51 may be constructed according to a semantic segmentation model based on deep learning technology. For example, a semantic segmentation model based on EfficientNetV3 and a semantic segmentation model based on the HRNet model may be adopted. It should be noted that EfficientNetV3 uniformly scales the depth, width, and resolution of the network, which can reduce the calculation and the number of parameters while ensuring the accuracy. And HRNet performs repeated multi-scale fusion by exchanging information between parallel multi-resolution sub-networks over and over again during the whole process, ensuring high resolution throughout the process, and having a good recognition effect on very small defects such as steel coil defects.

[0063] The semantic segmentation layer 51 can be used to classify each pixel in the target image into different semantic categories, such as the steel coil defect area and the non-defect area, so as to obtain semantic segmentation information. Among them, the semantic segmentation information can use different colors or grayscale values to represent different areas. Each pixel of a different semantic category has a different label. For example, the semantic label of the pixels in the steel coil defect area is 1, and the semantic label of the pixels in the non-defect area is 0.

[0064] In addition, the semantic segmentation layer 51 can be further used to identify the types of steel coil defect areas. In the semantic segmentation map, the pixels of various types of steel coil defect areas are identified by different labels, and different steel coil defect areas are displayed by different colors. The types of steel coil defect areas may include but are not limited to edge damage defects and edge crack defects.

[0065] The semantic segmentation layer 51 can also generate a semantic segmentation map. The semantic segmentation map is a visualization result of the semantic segmentation information.

[0066] The mask generation layer 52 can be used to represent the pixel values of the steel coil defect areas in the semantic segmentation information with mask values according to the semantic segmentation information output by the semantic segmentation layer 51, and represent all non-defect areas with pixel values that are at least 200 different from the mask values, so as to identify the mask information of the steel coil defect areas and the mask information of the non-defect areas.

[0067] The mask generation layer 52 can also generate a mask image based on the mask information. Among them, the mask image is a visualization result of the mask information. The mask image can be of various types. In one embodiment, the mask image is a binary image of the same size as the target image. In another embodiment, the mask image can be an image in which the steel coil defect areas in the target image are represented by mask values and the non-defect areas remain unchanged.

[0068] In some embodiments, the steel coil defect mask model can classify and store the recognition results of steel coil defects according to the types of steel coil defects. In this way, the search efficiency of the recognition results of specific types of steel coil defects can be improved, thereby improving the generation efficiency of specific type profiles.

[0069] In some embodiments, based on the above steel coil defect mask model. S21, perform semantic segmentation on the target image through the steel coil defect mask model to obtain the recognition result of the steel coil defect area, including: through the steel coil defect mask model, perform semantic segmentation and mask generation on the target image to obtain the mask information of the steel coil defect area.

[0070] In one embodiment, the recognition result of the steel coil defect area includes the mask information of the steel coil defect area. The mask information of the steel coil defect area is a special data representation generated for the steel coil defect area. Mask generation includes generating the mask of the steel coil defect area.

[0071] In another embodiment, in order to further distinguish the steel coil defect area from the non-defect area, mask generation further includes generating the mask of the non-defect area.

[0072] S22, based on the recognition result of the steel coil defect area, perform contour extraction on at least one steel coil defect area to obtain the contours of at least one steel coil defect, including: based on the mask information of the steel coil defect area, perform contour extraction on each steel coil defect area in the target image to obtain the contour of each steel coil defect.

[0073] In one embodiment, during the contour extraction process, use an image processing algorithm to process the mask information, identify the boundary of the steel coil defect area, and extract the contour of each steel coil defect, thereby obtaining the contour of each steel coil defect.

[0074] In the above embodiments, the boundary of the steel coil defect area can be more accurately determined through the mask information, and based on the mask information, the contour of the steel coil defect can be more accurately and quickly extracted, which is beneficial to improving the accuracy of steel coil defect recognition.

[0075] In some embodiments, after performing semantic segmentation and mask generation on the target image through the steel coil defect mask model to obtain the mask information of the steel coil defect area, the method further includes: generating a mask image based on the mask information of the steel coil defect area.

[0076] In one embodiment, when the mask information of the steel coil defect area is obtained in the mask generation layer 52, the mask image is an image of the same size as the target image. The difference between the mask image and the target image is that the steel coil defect area of the mask image is highlighted with a mask value.

[0077] In another embodiment, when the mask generation layer 52 obtains the mask information of the defective area and the non-defective area of the steel coil, the mask image is a binary image with the same size as the target image. The difference between the mask image and the target image is that the defective area and the non-defective area of the mask image are displayed with different mask values. Figure 6 and Figure 7 respectively show schematic diagrams of a mask image provided by an embodiment of the present application. As Figure 6 and Figure 7 shown, in the mask image, the candidate defective area of the steel coil is displayed in white, and the non-defective area is displayed in black. It should be noted that Figure 6 shows the edge loss defective area, Figure 7 shows the edge crack defective area.

[0078] In yet another embodiment, the non-defective area is divided into the normal area of the steel coil end face and the background area. When the mask generation layer 52 obtains the mask information of the defective area and the background area of the steel coil, the mask image is a binary image with the same size as the target image. The difference between the mask image and the target image is that the defective area and the background area of the mask image are displayed with different mask values.

[0079] In the above embodiment, by generating the mask image, the candidate defects of the steel coil can be made more prominent and easily recognizable in the image, thereby providing a basis for subsequent accurate contour extraction.

[0080] In some embodiments, through the steel coil defect mask model, semantic segmentation and mask generation are performed on the target image to obtain the mask information of the steel coil defect area, including: performing semantic segmentation on the target image through the semantic segmentation layer 51 of the steel coil defect mask model to obtain semantic segmentation information; through the mask generation layer 52 of the steel coil defect mask model, setting the pixel values in the steel coil defect area information to mask values to obtain the mask information of the steel coil defect area.

[0081] In one embodiment, the semantic segmentation information includes the steel coil defect area information and the non-defective area information. The steel coil defect area information includes the pixels belonging to each defect area, the position and range of each defect, and the corresponding semantic category label. The non-defective area information includes the pixels belonging to each non-defective area, the position and range of each non-defective area, and the corresponding semantic category label.

[0082] In one embodiment, the mask value is usually a specific numerical value used to highlight the defective area of the steel coil in the image. The semantic segmentation information is transmitted to the mask generation layer 52. The mask generation layer 52 is used to set the pixel values in the defect area information to a specific mask value to generate the mask information of the steel coil defect area.

[0083] In another embodiment, the mask generation layer 52 is further configured to set the pixel values in the defective area information to another specific mask value to generate the mask information for the background area.

[0084] In this way, through the specific mask value, the defective area of the steel coil can be clearly identified, forming an obvious difference from the non-defective area.

[0085] In some embodiments, the steel coil defect mask model sets the mask values for different types of defective areas according to the preset mask value rules. For example, the edge damage defective area corresponds to the first mask value, the edge crack defective area corresponds to the second mask value, and the non-defective area corresponds to the third mask value. Among them, the first mask value and the second mask value may be the same or different. The third mask value is different from the first mask value and also different from the second mask value. The difference between the third mask value and the first mask value or the second mask value is more than 200.

[0086] In one embodiment, the mask value includes the first mask value, and the steel coil defective area information includes the edge damage defective area information. Through the mask generation layer of the steel coil defect mask model, the pixel values in the steel coil defective area information are set to the mask value to obtain the mask information of the steel coil defective area, including: through the mask generation layer 52 of the steel coil defect mask model, the pixel values in the edge damage defective area information are set to the first mask value to obtain the mask information of the edge damage defective area in the target image.

[0087] In another embodiment, the mask value includes the second mask value, and the steel coil defective area information includes the edge crack defective area information. Through the mask generation layer of the steel coil defect mask model, the pixel values in the steel coil defective area information are set to the mask value to obtain the mask information of the steel coil defective area, including: through the mask generation layer 52 of the steel coil defect mask model, the pixel values of the edge crack defective area are set to the second mask value to obtain the mask information of the edge crack defective area in the target image.

[0088] In the above embodiments, by setting the mask values according to different defect types, the simultaneous detection and recognition of multiple defects can be achieved.

[0089] In yet another embodiment, through the mask generation layer of the steel coil defect mask model, the pixel values of the non-defective area can also be set to the third mask value to obtain the mask information of the non-defective area. And the mask image generated based on the mask information of the non-defective area and the mask information of the steel coil defective area can be as Figure 6 and Figure 7 shown. In this way, the obtained mask image can clarify the boundary between the steel coil defective area and the non-defective area, which is beneficial for users to understand the size and shape of the steel coil defects.

[0090] In some embodiments, through the mask generation layer 52 of the steel coil defect mask model, the pixel values of the background region can also be set to a third mask value to obtain the mask information of the background region. The mask generation layer 52 of the steel coil defect mask model does not mask the normal region of the steel coil end face, and the pixel values of the steel coil end face region remain unchanged.

[0091] In this way, the mask image generated based on the mask information of the non-defect region and the mask information of the steel coil defect region can prevent the user from being interfered by the background when observing the steel coil defects, and enable the user to quickly understand the distribution of the steel coil defects on the steel coil end face. Set the background region to correspond to the third mask value in the preset mask value rule.

[0092] In some embodiments, in order to comprehensively evaluate the quality of the steel coil, after obtaining the recognition result of the steel coil defect region by performing semantic segmentation on the target image through the steel coil defect mask model in S22, it further includes: obtaining the complete defect information of the steel coil end face based on the recognition result of the steel coil defect region of at least one target image.

[0093] Wherein, at least one target image includes a complete steel coil end face region, that is to say, one target image can correspond to a complete steel coil end face region, or multiple target images are integrated to obtain a complete steel coil end face region.

[0094] In one embodiment, multiple target images can be collected by the image acquisition device 11 when the steel coil rotates one week. Each target image contains different regions of the steel coil end face. Specifically, the steel coil defect recognition device 12 analyzes the steel coil defect information corresponding to different regions of the steel coil end face based on each target image, and then integrates the steel coil defect information in multiple target images to form the complete defect information of the steel coil end face.

[0095] In some embodiments, the steel coil recognition method may further include: after setting the pixel values of the background region to a third mask value through the mask generation layer 52 of the steel coil defect mask model to obtain the mask information of the background region, masking the background region in the target image based on the mask information of the background region to obtain a background-masked image.

[0096] S23, superimposing the contours of at least one steel coil defect onto the target image, including superimposing the contours of at least one steel coil defect onto the background-masked target image. Figure 8 A schematic diagram of another superimposed image provided by the embodiment of the present application is shown, as Figure 8 shown, the contours of the steel coil defects are marked by lines, and the background region is shown in black. In this way, the superimposed image can prevent the user from being interfered by the background and quickly observe the position and size of the steel coil defects on the steel coil end face.

[0097] In some embodiments, after calculating the bounding box for the contour of each steel coil defect to obtain the bounding box size of each steel coil defect, the steel coil defect recognition method may further include: superimposing the bounding box of each steel coil defect onto the target image.

[0098] Specifically, the steel coil defect recognition device 12 may superimpose the bounding box of each steel coil defect onto the target image according to the bounding box size of the bounding box of each steel coil defect. Among them, the bounding box may be represented by a rectangular frame. In this way, the image after superimposing the bounding box size should clearly show the position and size of the defect, enabling the user to intuitively see the actual size and position of the defect, thereby making decisions more easily.

[0099] In one embodiment, Figure 9 A schematic diagram showing a target image superimposing a contour and a bounding box provided by an embodiment of the present application is as Figure 9 shown. The bounding box of each steel coil defect may be superimposed with the contour in the original target image.

[0100] In another embodiment, Figure 10 A schematic diagram showing another target image superimposing a contour and a bounding box provided by an embodiment of the present application is as Figure 10 shown. The bounding box of each steel coil defect may be superimposed with the contour on the target image after the background mask at the same time. Compared with Figure 8 , Figure 10 the bounding box of the steel coil defect is added. The bounding box and the contour are represented by lines of different colors. It should be noted that Figure 10 the edge loss defect in

[0101] is relatively small, and most of the contour coincides with the bounding box. Figure 11 A schematic diagram showing the flow of another steel coil defect recognition method provided by an embodiment of the present application is as Figure 11 shown. The steel coil defect recognition method may include the following steps.

[0102] S1101, Obtain the target image.

[0103] S1102, Perform semantic segmentation and mask generation on the target image to obtain the mask information of the steel coil defect. Among them, the mask information of the steel coil defect may include the mask information of the edge loss defect and the mask information of the edge crack defect.

[0104] S1103, Extract the contour of each edge loss defect from the mask information of the edge loss defect.

[0105] S1104. Extract the contour of each edge crack defect from the mask information of the edge crack defect.

[0106] It should be noted that S1103 and S1104 are not in a sequential order and can be executed in parallel.

[0107] S1105. Perform bounding box calculation on the contour to obtain the maximum edge crack size and / or the maximum edge loss size.

[0108] S1106. Determine whether the maximum edge crack size exceeds the preset alarm edge crack size. If so, generate an edge crack alarm signal. If not, do not generate an edge crack alarm signal.

[0109] S1107. Determine whether the maximum edge loss size exceeds the preset alarm edge crack size. If so, generate an edge loss alarm signal. If not, do not generate an edge loss alarm signal.

[0110] Another aspect of the embodiments of the present application provides a steel coil defect recognition device. Figure 12 The structural schematic diagram of a steel coil defect recognition device provided by the embodiments of the present application is shown. As Figure 12 shown, the steel coil defect recognition device 120 may include the following several modules.

[0111] The semantic segmentation module 121 is used to perform semantic segmentation on the target image through the steel coil defect mask model to obtain the recognition result of the steel coil defect area, where the target image is an image including at least part of the end face of the steel coil.

[0112] The contour extraction module 122 is used to extract the contour of the steel coil defect area in the target image based on the recognition result of the steel coil defect area to obtain the contour of at least one steel coil defect.

[0113] The calculation module 123 is used to perform bounding box calculation on the contour of each steel coil defect to obtain the bounding box size of each steel coil defect.

[0114] The alarm signal generation determination module 124 is used to determine whether to generate a defect alarm signal according to the comparison result between the maximum bounding box size in the bounding box sizes of the at least one steel coil defect and the preset alarm size.

[0115] The contour superposition module 125 is used to superpose the contours of at least one steel coil defect onto the target image.

[0116] In the above embodiments, semantic segmentation is performed on the target image through the steel coil defect mask model to obtain the recognition result of the steel coil defect area. Thus, by capturing the local features of the steel coil end face, the detailed features of the steel coil end face can be expanded, and the steel coil defect mask model can be used to efficiently and accurately identify the defect area in the steel coil image. Based on the recognition result of the steel coil defect area, the contour of the steel coil defect area in the target image is extracted to obtain the contour of at least one steel coil defect. In this way, the accurate information of the steel coil defect can be obtained by extracting the contour of the steel coil defect area. By performing bounding box calculation on the contour of each steel coil defect, the position and size of the steel coil defect in the target image can be accurately determined, thereby further accurately measuring the size of the defect. In this way, the shape and size of the defect can be effectively described. And using the bounding box size as the basis for measuring the size of the defect enables defects of different shapes and different directions to be described and compared with a unified size parameter, which is conducive to establishing a unified defect alarm standard and making the alarm result more objective. By comparing the maximum bounding box size with the preset alarm size, it is possible to more accurately determine whether a defect alarm signal is generated, which helps to reduce false alarms, improve the reliability and effectiveness of the alarm signal, as well as improve the alarm efficiency and alarm sensitivity. At the same time, the defect alarm signal can effectively prompt the user in a timely manner that there are steel coil defects that seriously affect the quality of the steel coil. Overlaying the contour of at least one steel coil defect onto the target image can help to more intuitively display the specific position and shape of the defect, and facilitate the user to confirm in a timely and accurate manner the steel coil defects existing in the local area of the steel coil end face.

[0117] In some embodiments, the target bounding box size includes the maximum bounding box size. The alarm signal generation determination module 124 is specifically configured to determine whether to generate a defect alarm signal according to the comparison result between the maximum bounding box size among the bounding box sizes of at least one steel coil defect and the preset alarm size.

[0118] In some embodiments, the alarm signal generation determination module 124 is further specifically configured to generate a defect alarm signal when the maximum bounding box size among the bounding box sizes of at least one steel coil defect is greater than the preset alarm size; and determine not to generate a defect alarm signal when the maximum bounding box size among the bounding box sizes of at least one steel coil defect is not greater than the preset alarm size.

[0119] In some embodiments, the alarm signal generation determination module 124 includes a side damage alarm signal generation determination sub-module.

[0120] The side damage alarm signal generation determination sub-module is configured to determine whether to generate a side damage alarm signal according to the comparison result between the maximum bounding box size among the bounding box sizes of at least one side damage defect and the preset alarm side damage size when at least one steel coil defect includes at least one side damage defect.

[0121] In some embodiments, the alarm signal generation determination module 124 includes an edge crack alarm signal generation determination sub-module.

[0122] The edge crack alarm signal generation determination sub-module is configured to determine whether to generate an edge crack alarm signal according to the comparison result between the maximum bounding box size among the bounding box sizes of the at least one edge crack defect and a preset alarm edge crack size when the at least one steel coil defect includes at least one edge crack defect.

[0123] In some embodiments, the semantic segmentation module 121 is specifically configured to perform semantic segmentation and mask generation on a target image through a steel coil defect mask model to obtain mask information of the steel coil defect region.

[0124] The contour extraction module 122 is specifically configured to perform contour extraction on each steel coil defect region in the target image based on the mask information of the steel coil defect region to obtain the contour of each steel coil defect.

[0125] In some embodiments, the semantic segmentation module 121 may include a semantic segmentation sub-module and a mask generation sub-module.

[0126] The semantic segmentation sub-module is configured to perform semantic segmentation on the target image through the semantic segmentation layer of the steel coil defect mask model to obtain semantic segmentation information, where the semantic segmentation information includes steel coil defect region information and non-defect region information.

[0127] The mask generation sub-module is configured to set the pixel values in the steel coil defect region information to mask values through the mask generation layer of the steel coil defect mask model to obtain the mask information of the steel coil defect region.

[0128] In some embodiments, the mask value includes a first mask value. The mask generation sub-module includes a first mask generation sub-unit.

[0129] The first mask generation sub-unit is configured to set the pixel values in the edge damage defect region information to the first mask value through the mask generation layer of the steel coil defect mask model to obtain the mask information of the edge damage defect region in the target image.

[0130] In some embodiments, the mask value includes a second mask value. The mask generation sub-module includes a second mask generation sub-unit.

[0131] The second mask generation sub-unit is configured to set the pixel values of the edge crack defect region to the second mask value through the mask generation layer of the steel coil defect mask model to obtain the mask information of the edge crack defect region in the target image.

[0132] In some embodiments, the steel coil defect recognition device 120 may further include a mask image generation module.

[0133] The mask image generation module is configured to generate a mask image based on the mask information after performing semantic segmentation and mask generation on the target image through the steel coil defect mask model to obtain the mask information of the steel coil defect area.

[0134] In some embodiments, the steel coil defect recognition device 120 may further include a complete defect information acquisition module.

[0135] The complete defect information acquisition module is configured to obtain the complete defect information of the steel coil end face based on the recognition results of the steel coil defect areas of at least one target image, where the at least one target image includes a complete steel coil end face area.

[0136] It should be understood that the specific features, operations, and details described above regarding the method of the present application can also be similarly applied to the devices and systems of the present application, or vice versa. Additionally, each step of the method of the present application described above can be executed by the corresponding components or units of the device or system of the present application.

[0137] It should be understood that each module / unit of the device of the present application can be implemented in whole or in part by software, hardware, firmware, or a combination thereof. Each module / unit can be embedded in the processor of the electronic device in the form of hardware or firmware or independent of the processor, or stored in the memory of the electronic device in the form of software for the processor to call to execute the operations of each module / unit. Each module / unit can be implemented as an independent component or module, or two or more modules / units can be implemented as a single component or module.

[0138] Another aspect of the embodiments of the present application provides an electronic device. Figure 13 The structural schematic diagram of an electronic device provided by the embodiments of the present application is shown. As Figure 13 shown, the electronic device 130 includes a processor 131 and a memory 132 storing computer program instructions. Among them, when the processor 131 executes the computer program instructions, each step of the above-mentioned steel coil defect recognition method is implemented. The electronic device 130 can generally be a server, a terminal, or any other electronic device with necessary computing and / or processing capabilities.

[0139] In one embodiment, the electronic device 130 may include a processor, a memory, a network interface, a communication interface, etc. connected via a system bus. The processor of the electronic device 130 may be used to provide necessary computing, processing, and / or control capabilities. The memory of the electronic device 130 may include a non-volatile storage medium and an internal memory. The non-volatile storage medium may store an operating system, computer programs, etc. The internal memory may provide an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface and communication interface of the electronic device 130 may be used to connect and communicate with external devices via a network. When the computer program is executed by the processor, it performs the steps of the image edge detection method of the present application.

[0140] The present application provides a computer-readable storage medium, on which a steel coil defect identification program instruction is stored. When the steel coil defect alarm program instruction is executed by a processor, it implements the above-mentioned steel coil defect identification method.

[0141] Those skilled in the art can understand that the method steps of the present application can be instructed by a computer program to complete relevant hardware such as the electronic device 130 or the processor. The computer program can be stored in a non-transitory computer-readable storage medium. When the computer program is executed, it causes the steps of the present application to be executed. Depending on the situation, any reference to a memory, storage, or other medium herein may include non-volatile or volatile memory. Examples of non-volatile memory include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid state disk, etc. Examples of volatile memory include random access memory (RAM), external cache memory, etc.

[0142] The technical features described above can be combined arbitrarily. Although not all possible combinations of these technical features are described, any combination of these technical features should be considered to be covered by this specification as long as such a combination does not exist in contradiction.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for identifying steel coil defects, characterized in that, Including: Performing semantic segmentation on a target image through a steel coil defect mask model to obtain an identification result of the steel coil defect area, where the target image is an image including at least part of the end face of the steel coil; Based on the identification result of the steel coil defect area, extracting the contour of the steel coil defect area in the target image to obtain the contour of at least one steel coil defect; Calculating the bounding box for the contour of each steel coil defect to obtain the bounding box size of each steel coil defect; Determining whether to generate a defect alarm signal according to the comparison result between the maximum bounding box size among the bounding box sizes of the at least one steel coil defect and a preset alarm size; Superimposing the contour of the at least one steel coil defect onto the target image; Wherein, the steel coil defects include edge damage defects and edge crack defects, and the bounding box size includes the width and height of the bounding box.

2. The method for identifying steel coil defects according to claim 1, wherein The determining whether to generate a defect alarm signal according to the comparison result between the maximum bounding box size among the bounding box sizes of the at least one steel coil defect and a preset alarm size includes: Generating a defect alarm signal when the maximum bounding box size among the bounding box sizes of the at least one steel coil defect is greater than the preset alarm size; Determining not to generate a defect alarm signal when the maximum bounding box size among the bounding box sizes of the at least one steel coil defect is not greater than the preset alarm size.

3. The steel coil defect recognition method according to claim 1 or 2, characterized in that, The determining whether to generate a defect alarm signal according to the comparison result between the maximum bounding box size among the bounding box sizes of the at least one steel coil defect and a preset alarm size includes: When the at least one steel coil defect includes at least one edge damage defect, determining whether to generate an edge damage alarm signal according to the comparison result between the maximum bounding box size among the bounding box sizes of the at least one edge damage defect and a preset alarm edge damage size.

4. The method for identifying steel coil defects according to claim 1 or 2, characterized in that, The determining whether to generate a defect alarm signal according to the comparison result between the maximum bounding box size among the bounding box sizes of the at least one steel coil defect and a preset alarm size includes: When the at least one steel coil defect includes at least one edge crack defect, determining whether to generate an edge crack alarm signal according to the comparison result between the maximum bounding box size among the bounding box sizes of the at least one edge crack defect and a preset alarm edge crack size.

5. The steel coil defect recognition method according to claim 1, characterized in that The performing semantic segmentation on a target image through a steel coil defect mask model to obtain an identification result of the steel coil defect area includes: Performing semantic segmentation and mask generation on the target image through the steel coil defect mask model to obtain mask information of the steel coil defect area; The extracting the contour of the steel coil defect area in the target image to obtain the contour of at least one steel coil defect based on the identification result of the steel coil defect area includes: Based on the mask information of the steel coil defect area, extracting the contour of each steel coil defect area in the target image to obtain the contour of each steel coil defect.

6. The steel coil defect recognition method according to claim 5, characterized in that, The performing semantic segmentation and mask generation on a target image through the steel coil defect mask model to obtain the mask information of the steel coil defect area includes: Through the semantic segmentation layer of the steel coil defect mask model, semantic segmentation is performed on the target image to obtain semantic segmentation information, where the semantic segmentation information includes steel coil defect area information and non-defect area information; Through the mask generation layer of the steel coil defect mask model, the pixel values in the steel coil defect area information are set to mask values to obtain the mask information of the steel coil defect area.

7. The steel coil defect recognition method according to claim 6, wherein The steel coil defect area information includes edge damage defect area information. The step of, through the mask generation layer of the steel coil defect mask model, setting the pixel values in the steel coil defect area information to mask values to obtain the mask information of the steel coil defect area includes: Through the mask generation layer of the steel coil defect mask model, setting the pixel values in the edge damage defect area information to a first mask value to obtain the mask information of the edge damage defect area in the target image; and / or Through the mask generation layer of the steel coil defect mask model, setting the pixel values of the edge crack defect area to a second mask value to obtain the mask information of the edge crack defect area in the target image.

8. The method for identifying steel coil defects according to claim 5, characterized in that, After performing semantic segmentation and mask generation on the target image through the steel coil defect mask model to obtain the mask information of the steel coil defect area, it further includes: Generating a mask image based on the mask information.

9. The steel coil defect identification method according to claim 1, wherein After performing semantic segmentation on the target image through the steel coil defect mask model to obtain the recognition result of the steel coil defect area, it further includes: Based on the recognition results of the steel coil defect areas of at least one target image, obtaining the complete defect information of the steel coil end face, where the at least one target image includes a complete steel coil end face area.

10. A steel coil defect identification device, characterized in that, It includes: A semantic segmentation module for performing semantic segmentation on a target image through a steel coil defect mask model to obtain the recognition result of the steel coil defect area, where the target image is an image including at least part of the steel coil end face; A contour extraction module for extracting the contours of the steel coil defect areas in the target image based on the recognition result of the steel coil defect area to obtain the contours of at least one steel coil defect; A resolution module for performing bounding box resolution on the contours of each steel coil defect to obtain the bounding box sizes of each steel coil defect; An alarm signal generation determination module for determining whether to generate a defect alarm signal according to the comparison result between the maximum bounding box size among the bounding box sizes of the at least one steel coil defect and a preset alarm size; A contour superposition module for superposing the contours of the at least one steel coil defect onto the target image; Wherein, the steel coil defects include edge damage defects and edge crack defects, and the bounding box sizes include the width and height of the bounding box.

Images