Method and system for detecting shedding performance of zinc coating

Abstract

The invention discloses a zinc coating shedding performance detection method and system, particularly relates to the technical field of metal surface treatment quality detection, and is used for solving the problems of high misjudgment rate and insufficient detection precision caused by coating surface pseudo defect interference in the existing method. The method comprises the following steps: synchronously acquiring a multispectral reflection image and three-dimensional morphology data of the surface of a zinc coating through an industrial vision imaging device, extracting grain boundary distribution information based on multispectral data, and identifying radial pseudo defects in combination with a curvature manifold geodesic line energy gradient direction of the three-dimensional morphology; and further performing micro-crack topology analysis and grain boundary spatial correlation verification on the candidate region, eliminating texture interference and grain boundary overlapping regions, and finally outputting a high-confidence fall-off defect detection result, so that the identification capability of real defects under a complex galvanized surface is remarkably improved, and the method is suitable for automatic detection requirements of a high-speed continuous production line.

Description

Technical Field

[0001] The present invention relates to the technical field of metal surface treatment quality inspection, and more specifically, the present invention relates to a method and system for detecting the peeling performance of a galvanized layer. Background Art

[0002] In the field of metal material surface treatment, the galvanizing process is widely used to improve the corrosion resistance and mechanical properties of workpieces; in industrial scenarios, the detection of galvanized layer peeling defects is crucial for product quality control, and vision-based detection technology has become the mainstream method due to its non-contact and high-efficiency characteristics; existing methods usually collect images of the coating surface through an optical imaging system and identify abnormal areas based on image feature analysis to determine whether there are defects such as peeling and cracks; however, the galvanized layer surface often forms various non-defective features due to process parameter fluctuations, material composition differences, or environmental condition changes, such as textures formed by zinc liquid crystallization, spots generated by local oxidation reactions, and natural stress marks at the interface between the coating and the substrate; these features are highly similar to real peeling defects in terms of morphology, grayscale, or texture distribution, making it difficult for conventional detection means to effectively distinguish them.

[0003] Existing galvanized layer peeling detection methods have insufficient ability to identify surface pseudo-defects. Especially in a dynamic industrial environment, non-defective features generated by process conditions or material characteristics are easily misjudged as real peeling areas. Such misjudgments not only increase the cost of manual re-inspection but also lead to unnecessary downtime adjustments or material waste in the production process, seriously affecting the accuracy of detection results and the overall efficiency of the production line, which is particularly prominent in the detection of high-speed continuous galvanizing production lines or workpieces with complex surface structures. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method and system for detecting the peeling performance of a galvanized layer to solve the problems proposed in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for detecting the peeling performance of a galvanized layer, comprising the following steps:

[0007] S1. Collect multi-spectral reflection images and three-dimensional topography data of the galvanized layer surface, and extract galvanized layer grain boundary distribution data based on the multi-spectral reflection images;

[0008] S2. Extract candidate regions with abnormal reflection features based on the multi-spectral reflection images, and the candidate regions include zinc crystallization texture regions and peeling defect candidate regions;

[0009] S3. Extract the curvature manifold geodesic energy density gradient direction from the 3D topography data of the candidate regions, identify the pseudo-defect regions radially distributed in the zinc crystallization texture region, and exclude them;

[0010] S4. Conduct a microcrack topological structure analysis on the candidate regions of peeling defects after excluding the pseudo-defect regions, and extract the fracture crack propagation path features consistent with the substrate peeling direction;

[0011] S5. Compare the spatial correlation between the fracture crack propagation path features and the coating grain boundary distribution data, and exclude the pseudo-crack features that coincide with the grain boundary orientation;

[0012] S6. Output the detection results of galvanized coating peeling defects based on the fracture crack features regions not related to grain boundaries in the spatial correlation comparison results.

[0013] In a preferred embodiment, collect the multi-spectral reflection images and 3D topography data of the galvanized coating surface, and extract the coating grain boundary distribution data based on the multi-spectral reflection images, including:

[0014] Collect the multi-spectral reflection images of the galvanized coating surface through an industrial vision imaging device in the visible light band and the near-infrared band respectively;

[0015] Select adjacent pixel regions in the near-infrared band multi-spectral reflection image with a reflectivity difference exceeding a preset threshold, and mark them as grain boundary candidate regions;

[0016] Conduct a gray-scale gradient analysis on the visible light band multi-spectral reflection image, correct the misjudged regions caused by surface gloss interference in the grain boundary candidate regions, and generate the coating grain boundary distribution data;

[0017] Register the spatial coordinates of the coating grain boundary distribution data and the 3D topography data.

[0018] In a preferred embodiment, extract candidate regions with abnormal reflection features based on the multi-spectral reflection images. The candidate regions include the zinc crystallization texture region and the candidate regions of peeling defects, including:

[0019] Conduct a multi-band reflectivity ratio analysis on the multi-spectral reflection images, and calculate the reflectivity ratio map of the visible light band and the near-infrared band;

[0020] Extract the pixel regions with a reflectivity ratio lower than the first threshold in the reflectivity ratio map, and mark them as the zinc crystallization texture region;

[0021] Extract the pixel regions with a reflectivity ratio higher than the second threshold in the reflectivity ratio map, and conduct region growing in combination with the coating grain boundary distribution data to generate candidate regions of peeling defects;

[0022] Perform morphological closing operations on the zinc crystal texture region and the peeling defect candidate region, and merge adjacent pixels to form continuous candidate regions.

[0023] In a preferred embodiment, for the three-dimensional topography data of the candidate region, extract the curvature manifold geodesic energy density gradient direction, and identify and exclude the pseudo-defect regions radially distributed in the zinc crystal texture region, including:

[0024] Construct a curvature manifold for the three-dimensional topography data of the candidate region, and generate manifold space parameters based on the principal curvature direction and amplitude;

[0025] Generate geodesic paths in the curvature manifold, and calculate the energy density distribution along the geodesic paths;

[0026] Conduct gradient direction analysis on the energy density distribution, and extract the included angle data between the energy density gradient direction and the principal curvature direction;

[0027] Statistically analyze the distribution consistency of the included angle data within the zinc crystal texture region. If the included angle directions diverge radially, mark them as pseudo-defect regions;

[0028] Exclude the marked pseudo-defect regions from the candidate regions.

[0029] In a preferred embodiment, perform microcrack topological structure analysis on the peeling defect candidate region after excluding pseudo-defect regions, and extract the fracture crack propagation path characteristics consistent with the substrate peeling direction, including:

[0030] Perform binarization preprocessing on the peeling defect candidate region after excluding pseudo-defect regions to generate a microcrack connected domain skeleton diagram;

[0031] Extract the microcrack branch endpoints and intersection points based on the microcrack connected domain skeleton diagram, and construct a microcrack topological structure network;

[0032] Perform two-way path tracing along the microcrack branches in the microcrack topological structure network, and record the change amount of the included angle between the path propagation direction and the substrate peeling direction;

[0033] Select the microcrack branches with the continuously decreasing included angle between the path propagation direction and the substrate peeling direction, and extract the propagation path characteristics that meet the direction consistency;

[0034] Verify the curvature continuity of the propagation path characteristics that meet the direction consistency, and merge adjacent microcrack branches that pass the curvature continuity verification to form fracture crack propagation path characteristics.

[0035] In a preferred embodiment, perform spatial correlation comparison between the fracture crack propagation path characteristics and the coating grain boundary distribution data, and exclude the pseudo-crack characteristics that coincide with the grain boundary orientation, including:

[0036] Align the spatial coordinates of the fracture crack propagation path features with the distribution data of the coating grain boundaries to establish the position mapping relationship between the fracture crack propagation path features and the coating grain boundaries;

[0037] Calculate the Euclidean distance between each path point in the fracture crack propagation path features and the nearest grain boundary point in the coating grain boundary distribution data to generate a path-grain boundary distance distribution matrix;

[0038] In the path-grain boundary distance distribution matrix, filter out the path segments with Euclidean distance values less than the preset distance threshold and mark them as candidate pseudo-crack features;

[0039] Perform a direction similarity analysis on the direction of the candidate pseudo-crack features and the grain boundary directions in the coating grain boundary distribution data, and calculate the cosine similarity of their direction vectors;

[0040] Exclude the candidate pseudo-crack features with cosine similarity greater than the preset similarity threshold and update the fracture crack propagation path features.

[0041] In a preferred embodiment, according to the fracture crack feature regions not related to grain boundaries in the spatial correlation comparison result, output the galvanized coating peeling defect detection result, including:

[0042] Merge the fracture crack feature regions not related to grain boundaries in the spatial correlation comparison result to generate a continuous peeling defect region;

[0043] Based on the area and aspect ratio threshold of the continuous peeling defect region, determine the effective defect region;

[0044] Superimpose the effective defect region and the coating grain boundary distribution data spatially to generate a galvanized coating peeling defect detection result report;

[0045] According to the coordinate mapping relationship in the galvanized coating peeling defect detection result report, mark the position and range of the peeling defect in the three-dimensional topography data.

[0046] On the other hand, the present invention provides a galvanized coating peeling performance detection system, including the following modules:

[0047] Multi-spectrum analysis module: used to collect the multi-spectrum reflection images and three-dimensional topography data of the galvanized coating surface, and extract the coating grain boundary distribution data based on the multi-spectrum reflection images;

[0048] Feature selection module: used to extract candidate regions with abnormal reflection features based on the multi-spectrum reflection images, and the candidate regions include zinc crystal texture regions and peeling defect candidate regions;

[0049] Pseudo-defect filtering module: used to extract the curvature manifold geodesic energy density gradient direction from the three-dimensional topography data of the candidate regions, identify and exclude the pseudo-defect regions radially distributed in the zinc crystal texture regions;

[0050] Crack topology analysis module: used to perform micro-crack topology structure analysis on the peeling defect candidate areas after excluding pseudo-defect areas, and extract the characteristics of fracture crack propagation paths that are consistent with the peeling direction of the base material;

[0051] Grain boundary screening module: used to perform spatial correlation comparison between the characteristics of fracture crack propagation paths and the grain boundary distribution data of the coating, and exclude pseudo-crack characteristics that coincide with the grain boundary orientation;

[0052] Result output and display module: used to output the detection results of galvanized coating peeling defects according to the fracture crack feature areas that are not related to grain boundaries in the spatial correlation comparison results.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] 1. By integrating multi-spectral reflection imaging and three-dimensional topography analysis technologies, and combining a multi-dimensional verification mechanism of coating grain boundary distribution and micro-crack propagation path characteristics, the accuracy and reliability of galvanized coating peeling defect detection are significantly improved; By extracting grain boundary distribution data from multi-spectral reflection images and performing curvature manifold geodesic energy gradient analysis based on three-dimensional topography, radial pseudo-defects in zinc crystallization textures are effectively identified, solving the misjudgment problem caused by surface texture interference in traditional methods; At the same time, using micro-crack topology structure analysis and grain boundary spatial correlation comparison, pseudo-cracks that coincide with the grain boundary orientation are excluded from both the direction consistency and spatial distribution dimensions, ensuring that only real peeling defects are retained in the detection results;

[0055] 2. Through multi-modal data collaborative analysis and multi-level feature verification, accurate positioning of peeling defects under complex galvanized surfaces is achieved; Multi-spectral imaging provides information on material composition differences, and three-dimensional topography data captures surface deformation details. The combination of the two can suppress the influence of illumination changes and process noise; Based on the energy gradient direction screening of curvature manifold and grain boundary spatial correlation verification, a closed-loop logic from candidate area extraction to pseudo-defect exclusion is constructed, which is especially suitable for stable detection of dynamic surface defects in high-speed production lines. Description of the Drawings

[0056] Figure 1 It is a flowchart of a method for detecting the peeling performance of a galvanized coating according to the present invention;

[0057] Figure 2 It is a structural schematic diagram of a system for detecting the peeling performance of a galvanized coating according to the present invention. Detailed Embodiments

[0058] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0059] Embodiment 1: Figure 1 A method for detecting the peeling performance of a galvanized layer according to the present invention is provided, including the following steps:

[0060] S1. Collect multi-spectral reflection images and three-dimensional topography data of the galvanized layer surface, and extract the grain boundary distribution data of the coating based on the multi-spectral reflection images;

[0061] S2. Extract candidate regions with abnormal reflection characteristics based on the multi-spectral reflection images, and the candidate regions include zinc crystal texture regions and peeling defect candidate regions;

[0062] S3. For the three-dimensional topography data of the candidate regions, extract the curvature manifold geodesic energy density gradient direction, identify and exclude the pseudo-defect regions radially distributed in the zinc crystal texture regions;

[0063] S4. Conduct micro-crack topological structure analysis on the peeling defect candidate regions after excluding the pseudo-defect regions, and extract the fracture crack propagation path characteristics consistent with the substrate peeling direction;

[0064] S5. Perform spatial correlation comparison between the fracture crack propagation path characteristics and the grain boundary distribution data of the coating, and exclude the pseudo-crack characteristics that coincide with the grain boundary orientation;

[0065] S6. Output the detection result of the galvanized layer peeling defect according to the fracture crack feature regions not related to the grain boundary in the spatial correlation comparison result.

[0066] S1. Collect multi-spectral reflection images and three-dimensional topography data of the galvanized layer surface, and extract the grain boundary distribution data of the coating based on the multi-spectral reflection images. The specific implementation is as follows:

[0067] The industrial vision imaging device collects multi-spectral reflection images of the galvanized layer surface in the visible light band and the near-infrared band respectively, where the visible light band covers a wavelength range of 400 nm to 700 nm, and the near-infrared band covers a wavelength range of 800 nm to 1000 nm. The industrial vision imaging device uses a spectroscope prism combined with a filter wheel to achieve synchronous acquisition of multi-band images. The light source is an annular LED array, and the light incident direction forms a 45° angle with the normal of the coating surface to avoid specular reflection interference. The exposure time is dynamically adjusted according to the surface roughness of the coating. For example, when the surface roughness Ra value is greater than 1.6 μm, the exposure time is adjusted to 1.2 times the standard value to compensate for the light scattering loss.

[0068] Select adjacent pixel regions in the multi-spectral reflection image in the near-infrared band with reflectivity differences exceeding a preset threshold, and mark them as grain boundary candidate regions. The preset threshold is determined by statistically analyzing the reflectivity differences in the grain boundary regions of normal galvanized layer samples. For example, collect near-infrared reflection images of 50 defect-free samples, calculate the mean and standard deviation of the reflectivity differences between adjacent pixels, and set the threshold to the mean plus twice the standard deviation. When marking the grain boundary candidate regions, use the region growing algorithm. Take the pixels with reflectivity differences exceeding the threshold as seed points and expand to adjacent pixels until the reflectivity difference is lower than the threshold.

[0069] Perform gray-scale gradient analysis on the multi-spectral reflection image in the visible light band to correct the misjudged regions in the grain boundary candidate regions caused by surface gloss interference, and generate the coating grain boundary distribution data. The gray-scale gradient analysis includes calculating the gradient amplitude and direction of the gray-scale value change between a pixel point and its eight neighboring pixels. If the gradient amplitude in the visible light image of the grain boundary candidate region is less than 10 and the deviation of the gradient direction from the near-infrared reflectivity difference direction exceeds 30°, it is determined as a surface gloss interference region and excluded. For example, when the grain boundary candidate region marked in the near-infrared band shows a uniform gray-scale distribution and no directional change in the visible light image, this region is corrected as a misjudged region.

[0070] When registering the spatial coordinates of the coating grain boundary distribution data with the three-dimensional topography data, first project the feature points (such as grain boundary intersection points) in the three-dimensional topography data onto a two-dimensional plane to generate two-dimensional projection coordinates with the same viewing angle as the multi-spectral reflection image. The projection method uses orthogonal projection, and the projection plane is parallel to the imaging plane of the multi-spectral reflection image. For example, the coordinates of a grain boundary intersection point in the three-dimensional topography data are (x1, y1, z1), and the two-dimensional coordinates after projection are (x1', y1'); the coordinates of the corresponding point in the multi-spectral reflection image are (x2, y2). Optimize the parameters of the affine transformation matrix through the least squares method to minimize the spatial error between the projected two-dimensional coordinates (x1', y1') and the multi-spectral reflection image coordinates (x2, y2).

[0071] S2. Extract candidate regions with abnormal reflection characteristics based on the multi-spectral reflection image. The candidate regions include zinc crystal texture regions and shedding defect candidate regions. The specific implementation is as follows:

[0072] Perform multi - band reflectance ratio analysis on the multi - spectral reflection image to calculate the reflectance ratio map of the visible light band and the near - infrared band. The reflectance ratio is defined as the quotient of the reflectance value of the visible light band and the reflectance value of the near - infrared band. Among them, the green channel with a wavelength of 550nm is selected for the visible light band, and the channel with a wavelength of 900nm is selected for the near - infrared band. The green channel is sensitive to the oxide film in the zinc crystallization area, and the near - infrared channel is sensitive to the reflection characteristics of the zinc metal body. The ratio between the two can amplify the reflection difference between the zinc crystallization texture and the normal coating. When the industrial vision imaging device collects the multi - spectral reflection image, it is necessary to ensure that the imaging perspectives and lighting conditions of the two bands are consistent to avoid introducing errors in the ratio calculation. For example, when the green light reflectance of a certain pixel is 30% and the near - infrared reflectance is 60%, the reflectance ratio is 0.5.

[0073] Extract the pixel region with a reflectance ratio lower than the first threshold in the reflectance ratio map and mark it as the zinc crystallization texture region. The first threshold is determined by statistically analyzing the normal galvanized coating samples without texture interference. The specific method is as follows: Collect 30 galvanized coating samples without texture on the surface, calculate the reflectance ratio distribution of each sample, and take the mean of the reflectance ratios of all samples minus three times the standard deviation as the first threshold. For example, if the mean reflectance ratio of the normal samples is 0.8 and the standard deviation is 0.1, the first threshold is set to 0.5 (0.8 - 3×0.1). When marking the zinc crystallization texture region, use the connected - component analysis algorithm to merge adjacent pixels with a reflectance ratio lower than 0.5 into a continuous region.

[0074] Extract the pixel region with a reflectance ratio higher than the second threshold in the reflectance ratio map, and perform region growing in combination with the coating grain boundary distribution data to generate candidate regions for peeling defects. The second threshold is determined by statistically analyzing the reflectance ratios of known peeling defect samples. The specific method is as follows: Collect 20 galvanized coating samples with real peeling defects, calculate the reflectance ratio distribution of the defect regions, and take the mean plus twice the standard deviation as the second threshold. For example, if the mean reflectance ratio of the defect region is 1.2 and the standard deviation is 0.3, the second threshold is set to 1.8 (1.2 + 2×0.3). During the region - growing process, use the pixels with a reflectance ratio higher than 1.8 as seed points, and expand to adjacent pixels until the reflectance ratio is lower than 1.8. At the same time, exclude the regions marked as grain boundaries in the coating grain boundary distribution data. For example, when the reflectance ratio of the seed point A is 2.0 and the reflectance ratio of the adjacent pixel B is 1.7, stop expanding to B, and directly exclude B if it is within the grain boundary region.

[0075] The zinc crystal texture region and the candidate region for the detachment defect are processed by morphological closing operation, and the adjacent pixels are merged to form a continuous candidate region. The morphological closing operation uses a circular structure element with a diameter set to 5 pixels to fill the small holes inside the region and smooth the boundaries. For example, if there are discrete pixels caused by noise in the zinc crystal texture region (such as isolated points of 1-2 pixels), the closing operation can merge them with the adjacent region; if there are jagged bumps on the boundary of the candidate region for the detachment defect, the closing operation can smooth it into a continuous curve. The specific operations of the closing operation include expansion first and then corrosion. The expansion operation uses the structure element to expand the edge of the region, and the corrosion operation restores the main size of the region and removes burrs.

[0076] S3. Extract the energy density gradient direction of the geodesic curvature manifold from the three-dimensional morphology data of the candidate area, identify and exclude the pseudo-defect areas with radial distribution in the zinc crystal texture area, and implement it as follows:

[0077] The curvature manifold is constructed for the three-dimensional morphological data of the candidate area. The specific method is: based on the maximum principal curvature direction and the minimum principal curvature direction and their amplitude of each point in the three-dimensional morphological data, the spatial parameters of the curvature manifold are generated. The principal curvature direction indicates the maximum curvature direction of the surface at that point, and the principal curvature amplitude indicates the degree of curvature. For example, for the three-dimensional coordinates (x, y, z) of a certain point, the principal curvature direction (k1 direction and k2 direction) and amplitude (k1 and k2) are obtained by calculating the Hessi an matrix of its neighboring points. The curvature manifold maps the three-dimensional morphological data into a geometric space with the principal curvature as the core feature. The manifold space parameters include the principal curvature direction, amplitude and local surface normal vector. The calculation method of the Hessi an matrix is: perform secondary partial derivative calculations on the three-dimensional morphological data, construct a matrix containing second-order derivative information, and obtain the principal curvature direction and amplitude through eigenvalue decomposition.

[0078] A geodesic path is generated in a curvature manifold. The specific method is as follows: the center point of the zinc crystal texture area is selected as the starting point, and the geodesic path is extended along the main curvature direction. The geodesic path is the shortest path connecting two points on the manifold, and its energy density is defined as the change in the main curvature amplitude integrated along the path. For example, starting from the center point A and extending along the k1 direction to the edge point B, the change rate of the main curvature amplitude of each point on the path is calculated, and the energy density distribution is obtained by integration. The higher the energy density, the more drastic the curvature change on the path. The calculation of the energy density integral adopts the trapezoidal rule, discretizing the path into several line segments, and accumulating the curvature change segment by segment.

[0079] Perform a gradient direction analysis on the energy density distribution to extract the angle data between the energy density gradient direction and the principal curvature direction. The gradient direction represents the direction in which the energy density changes most rapidly. By calculating the angle between the gradient vector and the principal curvature direction vector, the energy diffusion pattern is judged. For example, if the angle between the gradient direction of a certain region and the k1 direction is 10°, it means that the energy diffuses uniformly along the principal curvature direction; if the angle is 80°, it means that the energy direction deviates greatly from the principal curvature direction. The calculation method of the gradient direction is as follows: perform a Sobel operator convolution on the energy density distribution, extract the gradient components in the horizontal and vertical directions, and synthesize the gradient vector field.

[0080] Statistically analyze the distribution consistency of the angle data of all points within the zinc crystallization texture region. If the angle directions show a radially divergent pattern, it is marked as a pseudo-defect region. The radially divergent pattern is defined as follows: with the texture center as the origin, the angle directions are evenly distributed along the radial direction and the angular dispersion is lower than a preset threshold. For example, for 100 points within a certain zinc crystallization texture region, calculate the standard deviation of their angle directions. If the standard deviation is less than 15°, it is determined to be a radial distribution. The standard deviation threshold is set by statistically analyzing the historical data of normal zinc crystallization texture samples. For example, collect 50 normal samples, calculate the mean value of the standard deviation of their angle directions as 12°, and set the threshold to 15°. The calculation method of the standard deviation is as follows: calculate the mean value of all angle direction values, calculate the mean value of the squared differences between each direction value and the mean value, and then take the square root.

[0081] Exclude the pseudo-defect regions marked as radially distributed from the candidate regions. The exclusion operation is achieved by updating the candidate region mask. For example, mark all pixels belonging to the pseudo-defect regions in the original candidate region as invalid, and only retain the remaining regions for subsequent analysis. The candidate regions after exclusion only contain real shedding defects and non-radial texture regions. The update method of the candidate region mask is as follows: apply the coordinate mapping relationship between the three-dimensional topography data and the multi-spectral reflection image to the mask marking to ensure spatial position consistency.

[0082] In step S3, by constructing a curvature manifold and analyzing the geodesic energy density gradient direction, it is possible to effectively distinguish between zinc crystallization texture pseudo-defects and real shedding defects. Compared with traditional two-dimensional image analysis or single curvature threshold segmentation methods, quantifying the energy diffusion characteristics of surface deformation based on three-dimensional topography data, combined with the radial growth law of zinc crystallization texture (an inherent process feature), can identify pseudo-defects caused by process fluctuations. Existing technologies usually ignore the correlation between three-dimensional topography and texture physical properties, resulting in a high misdetection rate in high-reflectivity or complex texture regions. Step S3 combines the geometric features of surface deformation and the energy diffusion pattern through curvature manifold mapping and gradient direction statistics, accurately distinguishing process artifacts (such as zinc flower crystallization) from real defects (such as crack peeling), improving the detection accuracy, and is especially suitable for stable detection under dynamic interference in high-speed production lines.

[0083] S4. Perform a microcrack topological structure analysis on the peeling defect candidate regions after excluding the pseudo-defect regions, and extract the characteristics of the fracture crack propagation path that is consistent with the substrate peeling direction. The specific implementation is as follows:

[0084] Perform binary preprocessing on the peeling defect candidate regions after excluding the pseudo-defect regions to generate a microcrack connected domain skeleton diagram. The binary preprocessing uses an adaptive threshold segmentation algorithm. The specific method is as follows: Calculate the mean and standard deviation of the pixel gray values in the candidate region, and set the threshold to the mean minus twice the standard deviation to enhance the contrast between the microcracks and the background. For example, if the gray mean of the candidate region is 120 and the standard deviation is 15, the threshold is 90 (120 - 2×15), and the pixels with gray values lower than 90 are determined to be the microcrack regions. After binaryzation, use a skeleton extraction algorithm to refine the microcrack connected domain to generate a microcrack connected domain skeleton diagram with a single-pixel width. The skeleton extraction algorithm retains the topological structure of the connected domain through iterative erosion operations, and each erosion operation removes the boundary pixels until the skeleton width is a single pixel.

[0085] Extract the microcrack branch endpoints and intersection points based on the microcrack connected domain skeleton diagram to construct a microcrack topological structure network. The microcrack branch endpoints are the isolated points at the ends of the skeleton lines, and the intersection points are the intersection points of the skeleton lines. The extraction method is as follows: Traverse each pixel in the skeleton diagram and count the number of eight-neighbor connected pixels. If the number is 1, it is an endpoint; if the number ≥ 3, it is an intersection point. For example, if a pixel has 2 connected pixels in its eight-neighborhood, it is determined to be an ordinary branch point; if it has 4 connected pixels, it is determined to be an intersection point. When constructing the microcrack topological structure network, use the endpoints and intersection points as network nodes, and the skeleton line branches as edges, and record the connection relationship between the nodes. After the network construction is completed, use an adjacency matrix to store the association information between the nodes and edges.

[0086] Perform two-way path tracking along the microcrack branches in the microcrack topological structure network, and record the change amount of the angle between the path extension direction and the substrate peeling direction. The two-way path tracking starts from an intersection point or an endpoint and extends along the skeleton line in two directions. The path extension direction is determined by calculating the coordinate change vector of adjacent pixel points, and the substrate peeling direction is set to be perpendicular to the normal direction of the coating surface according to the stress distribution at the interface between the galvanized layer and the substrate. For example, if the extension direction vector of a certain path is (1, 0) and the substrate peeling direction vector is (0, 1), the angle is 90°. During the tracking process, record the change amount of the angle every time a pixel is extended. If the angle continuously decreases, it is determined to be a path with consistent direction. The termination condition for path tracking is: reaching an endpoint or an intersection point, or the path length exceeding a preset threshold (for example, 50 pixels).

[0087] Screen for microcrack branches where the angle between the screening path extension direction and the substrate peeling direction continuously decreases, and extract the extension path features that meet the direction consistency. The determination condition for continuous decrease is that the difference in the angle change amount of three consecutive pixels during path extension is less than zero. For example, if the three consecutive angles of a path are 85°, 80°, and 75°, and the difference sequence is -5°, -5°, then it is determined to be continuously decreasing. The extension path features that meet the direction consistency need to satisfy that the decreasing trend of the angle covers more than 80% of the path length to ensure the stability of the peeling direction. If the path length is 10 pixels, then at least 8 pixel points' angles need to show a decreasing trend.

[0088] Perform curvature continuity verification on the extension path features that meet the direction consistency, and merge adjacent microcrack branches that pass the curvature continuity verification to form the fracture crack extension path features. The curvature continuity verification method is: calculate the curvature change rate of adjacent pixel points on the path. If the absolute value of the change rate is less than the preset threshold, it is determined to be continuous. The curvature is calculated through the coordinates of three adjacent points. For example, the curvature K of points A(x1, y1), B(x2, y2), and C(x3, y3) is determined by the change amount of the angle between vectors AB and BC. When merging adjacent branches, if the curvature change trends of the two branches are the same and the curvature difference is less than the threshold (for example, the curvature difference < 0.1), then they are merged into the same fracture crack extension path feature. The merging operation is realized by updating the adjacency matrix of the topological structure network, and the node association information of adjacent branches is marked as the same path.

[0089] In step S4, through microcrack topological structure analysis and direction consistency verification, the extension path features of real peeling defects can be accurately identified. Compared with the methods based on two-dimensional gray scale or simple morphological analysis in the prior art, this step combines three-dimensional topography data with the physical properties of crack propagation (such as the substrate peeling direction), and uses the topological network path tracking technology to quantify the correlation between the crack direction and the material stress distribution, so as to eliminate misjudgment caused by surface scratches or process textures. Due to the lack of modeling of the mechanical relationship between crack propagation direction and substrate peeling in the prior art, it is difficult to distinguish real peeling cracks from non-defective textures. Step S4 combines crack geometric features with material failure mechanisms through bidirectional path tracking and curvature continuity verification, making the detection results consistent with the actual peeling behavior of the coating, significantly improving the defect recognition accuracy under complex working conditions, and is especially suitable for the stable detection of microcracks in highly reflective galvanized coatings.

[0090] S5. Compare the spatial correlation between the fracture crack extension path features and the coating grain boundary distribution data, and exclude the pseudo-crack features that coincide with the grain boundary orientation. The specific implementation is as follows:

[0091] Align the spatial coordinates of the fracture crack propagation path features with the spatial coordinates of the coating grain boundary distribution data to establish the position mapping relationship between the fracture crack propagation path features and the coating grain boundaries. The coordinate alignment method is as follows: Based on the spatial coordinate mapping relationship between the registered coating grain boundary distribution data and the three-dimensional topography data in step S1, convert the three-dimensional coordinates of the fracture crack propagation path features into a two-dimensional coordinate system consistent with the coating grain boundary distribution data through an affine transformation matrix. For example, the three-dimensional coordinates of a path point in the fracture crack propagation path feature are (x, y, z), and the corresponding two-dimensional coordinates after affine transformation are (x', y'), which are in the same plane reference system as the grain boundary point coordinates in the coating grain boundary distribution data.

[0092] Calculate the Euclidean distance between each path point in the fracture crack propagation path feature and the nearest grain boundary point in the coating grain boundary distribution data to generate a path-grain boundary distance distribution matrix. The Euclidean distance calculation uses the two-dimensional plane geometric distance formula. When calculating, traverse all path points of the fracture crack propagation path feature, search for the nearest grain boundary point in the coating grain boundary distribution data for each path point, and record its Euclidean distance value. For example, if the distance between a path point and the nearest grain boundary point is 0.05 mm, record this value at the corresponding position in the path-grain boundary distance distribution matrix. The row index of the matrix corresponds to the path point number, and the column index corresponds to the grain boundary point number.

[0093] Screen out the path segments with Euclidean distance values less than the preset distance threshold in the path-grain boundary distance distribution matrix and mark them as candidate pseudo-crack features. The preset distance threshold is set according to the average physical width of the galvanized coating grain boundaries. For example, when the average grain boundary width is 0.1 mm, the threshold is set to 0.15 mm (1.5 times the grain boundary width). The screening condition is that the Euclidean distances of three or more consecutive path points in the path segment are all less than the threshold. For example, a path segment contains five consecutive points, and their distance values are 0.12 mm, 0.13 mm, 0.14 mm, 0.12 mm, and 0.11 mm, all less than 0.15 mm, then this path segment is marked as a candidate pseudo-crack feature.

[0094] Conduct a direction similarity analysis on the trend of the candidate pseudo-crack features and the grain boundary trend in the coating grain boundary distribution data, and calculate the cosine similarity of their direction vectors. The method for extracting the direction vector is as follows: Perform a least squares linear fitting on the path segment of the candidate pseudo-crack feature to obtain its main direction vector; perform the same fitting process on the corresponding grain boundary segment to obtain the grain boundary direction vector.

[0095] The cosine similarity calculation formula is cosθ = (A·B) / (||A||·||B||), where A and B are the coordinate components of the two vectors. For example, if the direction vector of the candidate pseudo-crack feature is (1, 0.5) and the grain boundary direction vector is (0.8, 0.4), then the cosine similarity is 0.992, which is determined to be highly similar.

[0096] Exclude candidate pseudo-crack features with a cosine similarity greater than the preset similarity threshold, and update the fracture crack propagation path features. The preset similarity threshold is set based on the statistical distribution of the angles between normal grain boundaries and real cracks. For example, 100 groups of samples of normal grain boundaries and real cracks are collected, and their average cosine similarity is calculated to be 0.2. The threshold is set to 0.9 to cover the process fluctuation range. If the cosine similarity of the candidate pseudo-crack feature exceeds 0.9, it is determined as a pseudo-crack feature that coincides with the grain boundary trend and is excluded. The exclusion operation is achieved by updating the validity flag of the fracture crack propagation path features. For example, the path segments that meet the exclusion conditions are removed from the list of valid features, and the remaining path segments are retained for the output of the final detection result.

[0097] In step S5, through spatial correlation comparison and direction similarity analysis, the misjudgment of pseudo-cracks caused by grain boundary interference can be accurately excluded. Compared with the existing methods that only rely on a single distance threshold or simple morphological matching, this step combines the spatial coordinate alignment of the fracture crack propagation path and the grain boundary, Euclidean distance statistics, and direction vector similarity verification to screen pseudo-features from the dual dimensions of position proximity and trend consistency. Through the joint determination of the distance threshold (covering the grain boundary diffusion area) and the cosine similarity threshold (quantifying the trend consistency), the exclusion rate of grain boundary artifacts is improved, ensuring that the detection result strictly reflects the actual peeling state of the coating, especially suitable for defect identification in complex grain boundary backgrounds in high-precision industrial detection scenarios.

[0098] S6. Output the detection result of the galvanized layer peeling defect according to the fracture crack feature region not related to the grain boundary in the spatial correlation comparison result. The specific implementation is as follows:

[0099] Merge the fracture crack feature regions not related to the grain boundary in the spatial correlation comparison result to generate a continuous peeling defect region. The region merging adopts the morphological closing operation algorithm. The specific method is as follows: Select a circular structural element with a diameter set to 5 pixels, perform a dilation operation on the fracture crack feature region not related to the grain boundary to fill the region gap, and then perform an erosion operation to restore the original size of the region. For example, if the distance between two adjacent fracture crack feature regions is less than 3 pixels, the closing operation can merge them into a continuous region. The merged continuous peeling defect region needs to meet the minimum area threshold (for example, 0.1mm 2 ) to exclude isolated small regions generated by noise interference.

[0100] Based on the area and aspect ratio threshold of the continuous peeling defect region, determine the effective defect region. The area threshold is set according to the minimum physical size of the galvanized layer peeling defect. For example, in the actual process, the lower limit of the recognizable defect area is 0.5mm 2 , then the threshold is set to 0.5mm 2。The aspect ratio threshold is used to exclude non-shedding defects such as linear scratches, and is specifically set to 3:1 (length / width). For example, the area of a continuous shedding defect region is 0.6 mm 2 , and the aspect ratio is 2.5:1, then it is determined as a valid defect region; if the area of another region is 0.4 mm 2 or the aspect ratio is 4:1, it is excluded. During the determination process, defect regions with an overlap rate exceeding 5% with the grain boundary regions in the grain boundary distribution data of the coating are also excluded to ensure that the detection results are not interfered by the grain boundaries.

[0101] The valid defect regions are spatially superimposed with the grain boundary distribution data of the coating to generate a detection result report on the shedding defects of the galvanized coating. The spatial superimposition method is as follows: Align the coordinates of the valid defect regions with the coordinates of the grain boundary distribution data of the coating in the same coordinate system, and mark the position, area, and shape parameters of the defect regions. The detection result report is presented in the form of a combination of a two-dimensional image and three-dimensional topography data. For example, the defect boundary is marked with a red polygon in the two-dimensional image, and the defect depth distribution is marked with contour lines in the three-dimensional topography data.

[0102] Based on the coordinate mapping relationship in the detection result report on the shedding defects of the galvanized coating, mark the position and range of the shedding defects in the three-dimensional topography data. The coordinate mapping relationship is realized based on the registration result of the grain boundary distribution data and the three-dimensional topography data in step S1. For example, the two-dimensional coordinates of a certain defect in the detection result report are (x1, y1), which are mapped to the coordinates (x1', y1', z1) in the three-dimensional topography data through an affine transformation matrix, and the height value z1 is extracted from the three-dimensional point cloud data to generate three-dimensional annotation information of the defect including the position (x1', y1', z1) and the range (length, width, depth).

[0103] Embodiment 2: Figure 2 The structural schematic diagram of a detection system for the shedding performance of a galvanized coating according to the present invention is given. A detection system for the shedding performance of a galvanized coating includes the following modules:

[0104] Multi-spectrum acquisition and analysis module: used to acquire the multi-spectrum reflection image and three-dimensional topography data of the surface of the galvanized coating, and extract the grain boundary distribution data of the coating based on the multi-spectrum reflection image;

[0105] Feature selection module: used to extract candidate regions with abnormal reflection features based on the multi-spectrum reflection image. The candidate regions include zinc crystal texture regions and shedding defect candidate regions;

[0106] Pseudo-defect filtering module: used to extract the curvature manifold geodesic energy density gradient direction from the three-dimensional topography data of the candidate regions, identify and exclude the pseudo-defect regions radially distributed in the zinc crystal texture regions;

[0107] Crack Topology Analysis Module: It is used to perform micro-crack topology structure analysis on the candidate areas of peeling defects after excluding pseudo-defect areas, and extract the characteristics of the fracture crack propagation path consistent with the peeling direction of the substrate;

[0108] Grain Boundary Screening Module: It is used to perform a spatial correlation comparison between the characteristics of the fracture crack propagation path and the grain boundary distribution data of the coating, and exclude the pseudo-crack characteristics that coincide with the grain boundary orientation;

[0109] Result Output and Display Module: It is used to output the detection result of the galvanized coating peeling defect according to the fracture crack feature area not related to the grain boundary in the result of the spatial correlation comparison.

[0110] The calculations involved in the embodiments are all numerical calculations after removing the dimensions. The preset parameters and threshold selection in the calculations are set by those skilled in the art according to the actual situation.

[0111] It should be noted that the present invention can be deployed on the device itself to achieve embedded applications, or can also run on a PC or other terminals with a user interface, so as to meet various hardware environments and usage requirements.

[0112] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on the computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wire (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server or data center that includes one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0113] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described systems, devices, and modules can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0114] In several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division. In actual implementation, there can be other division methods. For example, multiple modules or components can be combined or integrated into another system, 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 interfaces. The indirect couplings or communication connections of devices or modules can be in electrical, mechanical, or other forms.

[0115] The modules described as separate components may or may not be physically separated. The components shown as modules may or may not be physical modules. They can be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0116] In addition, in each embodiment of this application, the functional modules can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0117] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of this application. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs and other various media that can store program codes.

[0118] As described above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0119] Finally, the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for detecting the peeling performance of a galvanized layer, characterized in that, It includes the following steps: S1. Collect the multi-spectral reflection images and three-dimensional topography data of the galvanized layer surface, and extract the distribution data of the coating grain boundaries based on the multi-spectral reflection images; S2. Extract candidate regions with abnormal reflection characteristics based on the multi-spectral reflection images. The candidate regions include zinc crystal texture regions and exfoliation defect candidate regions; S3. For the three-dimensional topography data of the candidate regions, extract the curvature manifold geodesic energy density gradient direction, identify the pseudo-defect regions radially distributed in the zinc crystal texture regions and exclude them; S4. Conduct a micro-crack topological structure analysis on the exfoliation defect candidate regions after excluding the pseudo-defect regions, and extract the fracture crack propagation path characteristics consistent with the substrate peeling direction; S5. Conduct a spatial correlation comparison between the fracture crack propagation path characteristics and the coating grain boundary distribution data, and exclude the pseudo-crack characteristics that coincide with the grain boundary trend; S6. Output the detection results of the galvanized layer exfoliation defects according to the fracture crack characteristics regions not related to the grain boundaries in the spatial correlation comparison results.

2. The galvanized layer peeling performance detection method according to claim 1, wherein Collect the multi-spectral reflection images and three-dimensional topography data of the galvanized layer surface, and extract the distribution data of the coating grain boundaries, including: Collect the multi-spectral reflection images of the galvanized layer surface through an industrial vision imaging device in the visible light band and the near-infrared band respectively; Select adjacent pixel regions with a reflectance difference exceeding a preset threshold in the near-infrared band multi-spectral reflection image and mark them as grain boundary candidate regions; Conduct a gray gradient analysis on the visible light band multi-spectral reflection image, correct the misjudged regions in the grain boundary candidate regions caused by surface gloss interference, and generate the coating grain boundary distribution data; Register the spatial coordinates of the coating grain boundary distribution data and the three-dimensional topography data.

3. The galvanized layer peeling performance detection method according to claim 1, characterized in that, Extract candidate regions with abnormal reflection characteristics based on the multi-spectral reflection images. The candidate regions include zinc crystal texture regions and exfoliation defect candidate regions, including: Conduct a multi-band reflectance ratio analysis on the multi-spectral reflection images to calculate the reflectance ratio map of the visible light band and the near-infrared band; Extract pixel regions with a reflectance ratio lower than the first threshold in the reflectance ratio map and mark them as zinc crystal texture regions; Extract pixel regions with a reflectance ratio higher than the second threshold in the reflectance ratio map, and conduct region growing in combination with the coating grain boundary distribution data to generate exfoliation defect candidate regions; Conduct a morphological closing operation on the zinc crystal texture regions and the exfoliation defect candidate regions, and merge adjacent pixels to form continuous candidate regions.

4. A method for detecting the peeling performance of a galvanized layer according to claim 1, characterized in that, For the three-dimensional topography data of the candidate regions, extract the curvature manifold geodesic energy density gradient direction, identify the pseudo-defect regions radially distributed in the zinc crystal texture regions and exclude them, including: Construct a curvature manifold for the three-dimensional topography data of the candidate regions, and generate manifold space parameters based on the principal curvature direction and amplitude; Generate geodesic paths in the curvature manifold and calculate the energy density distribution along the geodesic paths; Conduct a gradient direction analysis on the energy density distribution, and extract the included angle data between the energy density gradient direction and the principal curvature direction; Statistically analyze the distribution consistency of the included angle data in the zinc crystal texture regions. If the included angle directions diverge radially, mark them as pseudo-defect regions; Exclude the marked pseudo-defect regions from the candidate regions.

5. A method for detecting the peeling performance of a galvanized layer according to claim 1, characterized in that Perform microcrack topological structure analysis on the peeling defect candidate regions after excluding pseudo-defect regions, and extract the characteristics of the fracture crack propagation path consistent with the substrate peeling direction, including: Perform binary preprocessing on the peeling defect candidate regions after excluding pseudo-defect regions to generate a microcrack connected domain skeleton diagram; Extract microcrack branch endpoints and intersection points based on the microcrack connected domain skeleton diagram to construct a microcrack topological structure network; Perform two-way path tracing along the microcrack branches in the microcrack topological structure network, and record the change amount of the angle between the path propagation direction and the substrate peeling direction; Screen the microcrack branches with the continuously decreasing angle between the path propagation direction and the substrate peeling direction, and extract the expansion path characteristics that meet the direction consistency; Perform curvature continuity verification on the expansion path characteristics that meet the direction consistency, and merge adjacent microcrack branches that pass the curvature continuity verification to form fracture crack expansion path characteristics.

6. The galvanized layer peeling performance detection method according to claim 1, characterized in that Perform spatial correlation comparison between the fracture crack expansion path characteristics and the coating grain boundary distribution data, and exclude pseudo-crack characteristics that coincide with the grain boundary trend, including: Align the spatial coordinates of the fracture crack expansion path characteristics and the coating grain boundary distribution data to establish a position mapping relationship between the fracture crack expansion path characteristics and the coating grain boundaries; Calculate the Euclidean distance between each path point in the fracture crack expansion path characteristics and the nearest grain boundary point in the coating grain boundary distribution data to generate a path-grain boundary distance distribution matrix; Screen the path segments with Euclidean distance values less than the preset distance threshold in the path-grain boundary distance distribution matrix and mark them as candidate pseudo-crack characteristics; Perform direction similarity analysis on the trend of the candidate pseudo-crack characteristics and the grain boundary trend in the coating grain boundary distribution data, and calculate the cosine similarity of their direction vectors; Exclude the candidate pseudo-crack characteristics with cosine similarity greater than the preset similarity threshold, and update the fracture crack expansion path characteristics.

7. A method for detecting the peeling performance of a galvanized layer according to claim 1, characterized in that, According to the non-grain boundary-related fracture crack feature regions in the spatial correlation comparison results, output the galvanized layer peeling defect detection results, including: Merge the non-grain boundary-related fracture crack feature regions in the spatial correlation comparison results to generate continuous peeling defect regions; Based on the area and aspect ratio thresholds of the continuous peeling defect regions, determine the effective defect regions; Perform spatial superposition of the effective defect regions and the coating grain boundary distribution data to generate a galvanized layer peeling defect detection result report; According to the coordinate mapping relationship in the galvanized layer peeling defect detection result report, mark the position and range of the peeling defects in the three-dimensional topography data.

8. A galvanized layer peeling performance detection system for implementing the galvanized layer peeling performance detection method according to any one of claims 1-7, characterized in that, Including the following modules: Multi-spectrum analysis module: used to collect multi-spectral reflection images and three-dimensional topography data of the galvanized layer surface, and extract coating grain boundary distribution data based on the multi-spectral reflection images; Feature selection module: used to extract candidate regions with abnormal reflection characteristics based on the multi-spectral reflection images, and the candidate regions include zinc crystal texture regions and peeling defect candidate regions; Pseudo-defect filtering module: used to extract the curvature manifold geodesic energy density gradient direction from the three-dimensional topography data of the candidate regions, identify and exclude the pseudo-defect regions radially distributed in the zinc crystal texture regions; Crack topology analysis module: used to perform micro-crack topology structure analysis on the candidate areas of peeling defects after excluding pseudo-defect areas, and extract the characteristics of the fracture crack propagation path consistent with the peeling direction of the substrate; Grain boundary screening module: used to perform spatial correlation comparison between the fracture crack propagation path characteristics and the grain boundary distribution data of the coating, and exclude the pseudo-crack characteristics that coincide with the grain boundary trend; Result output and display module: used to output the detection results of galvanized coating peeling defects according to the fracture crack feature areas not related to grain boundaries in the spatial correlation comparison results.

Images