Automatic detection method for anti-corrosion coating quality of steel member based on unmanned aerial vehicle
By using drones to detect closed connected regions in images of steel components, and combining gradient differences and texture features, corrosion and flaking areas can be identified. This solves the problem of low accuracy in identifying corrosion and flaking areas in existing technologies, and enables more accurate inspection of anti-corrosion coating quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JSTI GRP CO LTD
- Filing Date
- 2026-01-30
- Publication Date
- 2026-05-12
AI Technical Summary
Existing technologies for inspecting the quality of anti-corrosion coatings on steel components are prone to overlooking or misidentifying areas of corrosion and peeling, resulting in low identification accuracy and affecting the inspection results.
An automated inspection method for the anti-corrosion coating quality of steel components based on UAVs was adopted. By analyzing the gradient differences, curvature distribution and texture features of edge pixels of closed connected domains and their local neighborhoods, the second abnormal feature value was determined, and the corrosion and peeling areas were identified.
It improves the accuracy of identifying corrosion-scraped areas, avoids misjudgments and omissions, enhances the robustness of detection, and ensures the accuracy of anti-corrosion coating quality inspection.
Smart Images
Figure CN122023342A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of image processing technology, specifically to an automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs). Background Technology
[0002] With the rapid development of the national economy and the continuous advancement of industrialization, steel structures, due to their advantages such as high strength, light weight, good seismic performance, and short construction period, are now widely used in many fields. Steel structures are commonly used in various engineering structures with large spans, large heights, large loads, and large dynamic effects. While steel structures are widely used, their quality assurance and service life have attracted considerable attention, and anti-corrosion coating is a key aspect in extending the lifespan of steel structures.
[0003] Anti-corrosion coatings for steel structures are crucial for ensuring the durability and safety of steel structural components, directly impacting their service life. In corrosive environments such as humidity, acid rain, salt spray, and industrial pollution, uncoated steel structures will rapidly rust, severely affecting their load-bearing capacity and potentially leading to safety accidents. By applying anti-corrosion coatings to steel components and ensuring the quality of the coating application, moisture, oxygen, and corrosive media can be effectively isolated, significantly reducing the likelihood of rust and extending the service life of the steel structures.
[0004] Currently, existing technologies typically employ machine vision to detect corrosion and flaking areas in images of steel components, thereby completing the quality inspection of the anti-corrosion coating on steel components. However, the shape, size, color, and texture of corrosion and flaking areas vary considerably. Early corrosion may only involve tiny spots, while severe corrosion can lead to large-area flaking. Different types of corrosion areas exhibit different characteristics, and existing detection algorithms are prone to ignoring or misidentifying some corrosion areas, affecting the accuracy of corrosion flaking area identification and reducing the effectiveness of anti-corrosion coating quality inspection on steel components. Summary of the Invention
[0005] To address the aforementioned technical problems, this application provides an automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs), thereby resolving the existing issues.
[0006] The automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs) of this application adopts the following technical solution: One embodiment of this application provides an automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs), the method comprising the following steps: Based on images of steel components acquired by drones, each closed connected component in the steel component image is obtained; Analyze the gradient difference between each edge pixel and its local neighborhood pixels in each closed connected domain to determine the first abnormal feature value of each edge pixel in each closed connected domain. Determine the randomness of the peak distribution in the curvature of all edge pixels in each closed connected domain, and combine the texture feature differences between each edge pixel and its local neighborhood pixels in each closed connected domain to determine the edge variability of each closed connected domain. By combining the edge volatility of each closed connected domain with the first abnormal feature value of all edge pixels in each closed connected domain, the second abnormal feature value of each closed connected domain is obtained. The corrosion and flaking areas in the steel component image are identified based on the second abnormal feature value.
[0007] In one embodiment, determining the first abnormal feature value includes: The average difference between the gradient magnitude of each edge pixel in each closed connected region and all pixels in its local neighborhood is calculated. Based on the difference between the average difference between each edge pixel in each closed connected region and the edge pixels in its local neighborhood, it is denoted as the first difference. The difference between the gradient angle of each edge pixel in each closed connected region and the pixel in its local neighborhood is used to determine the first abnormal feature value.
[0008] In one embodiment, the process of determining the first abnormal feature value is as follows: Calculate the difference in gradient angle between each edge pixel on each closed connected domain and each pixel in its local neighborhood, and calculate the square of the cosine of the difference. The first abnormal feature value is negatively correlated with the square of the cosine and positively correlated with the first difference.
[0009] In one embodiment, the expression for the first abnormal feature value is: In the formula, Let be the first anomaly feature value of the u-th edge pixel in each closed connected component, sig() be the sigmoid normalization function, and M be the number of pixels in the local neighborhood of the u-th edge pixel in each closed connected component. Let be the mean value of the u-th edge pixel of each closed connected component. Let be the mean value of the a-th pixel within the local neighborhood of the u-th edge pixel of each closed connected component, and Norm() be the normalization function. Let be the gradient angle of the u-th edge pixel in each closed connected region. Let be the gradient angle of the a-th pixel within the local neighborhood of the u-th edge pixel in each closed connected region, and let cos() be the trigonometric cosine function. The value is a preset value greater than 0, wherein when the a-th pixel in the local neighborhood of the u-th edge pixel is a non-edge pixel, the mean value of the a-th pixel is 0.
[0010] In one embodiment, determining the degree of randomness in the peak distribution of curvature of all edge pixels in each closed connected region includes: Arrange the curvature of all edge pixels in each closed connected region in a uniform order, obtain the peak value of the curvature of all edge pixels in each closed connected region after the arrangement, and determine the degree of randomness of the peak distribution of the curvature of all edge pixels in each closed connected region based on the number of curvatures between adjacent peak values.
[0011] In one embodiment, if the number of curvatures between any peak and its adjacent previous peak is equal to the number of curvatures between any peak and its adjacent next peak, the peak is marked as 0; otherwise, the peak is marked as 1. The marked values of all peaks of curvatures of all edge pixels of each closed connected domain are summed to determine the degree of randomness of the peak distribution in the curvatures of all edge pixels of each closed connected domain.
[0012] In one embodiment, determining the edge volatility of each closed connected component includes: The difference between the LBP values of each edge pixel and its local neighborhood pixels on each closed connected domain is calculated and denoted as the second difference. The fusion result of the second difference of all edge pixels on each closed connected domain is obtained. The edge volatility is positively correlated with the randomness and the fusion result.
[0013] In one embodiment, the edge volatility of each closed connected domain is the product of the randomness and the fusion result.
[0014] In one embodiment, determining the second abnormal feature value of each closed connected component includes: Calculate the mean of the first abnormal feature values of all edge pixels in each closed connected region, and multiply it by the normalized value of the edge volatility of each closed connected region to obtain the second abnormal feature value of each closed connected region.
[0015] In one embodiment, identifying corrosion and flaking areas in the steel component image based on the second abnormal feature value includes: Threshold segmentation is performed on the second abnormal feature value of all closed contours in the steel component image. If there is a closed contour in the steel component image with a second abnormal feature value greater than the segmentation threshold, it is determined that there is a corrosion and flaking area in the steel component image; otherwise, it is determined that there is no corrosion and flaking area in the steel component image.
[0016] This application has at least the following beneficial effects: This application analyzes closed connected regions in steel component images and combines the gradient differences between edge pixels of closed connected regions and pixels in their local neighborhoods to accurately determine the contours and boundaries of corrosion-scraped areas, ensuring correct identification of these areas. The gradient differences reflect significant edge changes, effectively identifying areas of material peeling caused by corrosion and avoiding interference from image quality degradation or complex backgrounds. Furthermore, determining the edge variability of each closed connected region accurately reflects the irregularity of corrosion areas on the steel component surface, improving the accuracy of corrosion-scraped area identification. The randomness of the peak distribution in the curvature of all edge pixels of closed connected regions further helps distinguish corrosion-scraped areas from other defect areas on the steel component surface, enhancing the detection capability for complex corrosion areas. The determination of the second abnormal feature value helps refine the identification of different corrosion levels on the steel component surface, more accurately distinguishing slightly corroded areas from background noise or other surface defects, avoiding misjudgments and omissions, improving the robustness of corrosion-scraped area identification in steel component images, and further enhancing the quality inspection effect of anti-corrosion coating on steel components. Attached Figure Description
[0017] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 A flowchart illustrating the steps of an automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs) provided in this application. Figure 2 Flowchart for determining the second abnormal feature value of a closed connected domain. Detailed Implementation
[0019] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs) proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.
[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.
[0021] The following description, in conjunction with the accompanying drawings, details the specific scheme of the automated inspection method for anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs) provided in this application.
[0022] This application provides an embodiment of an automated inspection method for the quality of anti-corrosion coatings on steel components based on unmanned aerial vehicles (UAVs). Specifically, it provides the following automated inspection method for the quality of anti-corrosion coatings on steel components based on unmanned aerial vehicles (UAVs). Please refer to [link to relevant documentation]. Figure 1 The method includes the following steps: Step S001: Based on the images of the steel components collected by the UAV, obtain each closed connected component in the images of the steel components.
[0023] To ensure the service life and safety of steel structure components, anti-corrosion coating is typically applied, and the coating quality must be inspected. In this embodiment, a professional drone is used to acquire image data of the anti-corrosion coating on the steel components during quality inspection. Specifically, a professional drone equipped with a 20-megapixel industrial camera is used to take parallel photos of the steel components, while a coaxial light source is used for illumination during the shooting process to avoid lighting interference during image acquisition.
[0024] After acquiring the steel component image, preprocessing is performed. To reduce subsequent computation, the acquired steel component image is converted to grayscale. Then, nonlocal means denoising is used to denoise the grayscale image, preserving edge details and enhancing the contrast of defects during steel component anti-corrosion coating quality inspection. Both the grayscale conversion and nonlocal means denoising of the steel component image are existing known techniques; implementers can choose other feasible denoising algorithms, and this embodiment does not impose any restrictions.
[0025] Given the widespread use of steel components in bridges, high-rise buildings, power towers, petrochemical facilities, ships, and airport terminals, their core function is to withstand high-intensity loads and provide stable structural support. These environments commonly experience high humidity, salt spray, and chemical corrosion, which can easily lead to coating aging and failure. Therefore, ensuring the quality of anti-corrosion coatings on steel components is crucial, requiring regular inspections to guarantee the safety and durability of the steel structure. However, when using drones to inspect the anti-corrosion coatings on steel components, corrosion and peeling may occur, potentially leading to quality problems with the steel components.
[0026] Since corrosion and flaking areas on steel components typically appear as patches or irregularly shaped regions, forming closed connected areas, the following steps are taken for quality inspection of the anti-corrosion coating on steel components. First, the surface area of the steel component is extracted from images acquired using a drone. The Canny edge detection algorithm is used to obtain the closed contour regions in the steel component images, i.e., to obtain each closed connected region in the steel component image. Considering that the drone images primarily focus on the surface of the steel component, and other interfering backgrounds account for a small proportion of the image, the acquired images mainly represent the surface of the steel component. Therefore, by calculating the area of all closed contours in the steel component image, the largest closed contour region is identified as the surface area to be inspected. Subsequent analysis is then performed on each closed connected region within this surface area. The Canny edge detection algorithm is a well-known existing technology, and its specific process will not be elaborated upon.
[0027] It should be noted that if there are no other closed connected regions on the surface of a steel component, it indicates that the anti-corrosion coating on the surface of the steel component is qualified and there are no areas of corrosion peeling.
[0028] Step S002: Analyze the gradient difference between each edge pixel and its local neighborhood pixels on each closed connected domain, and determine the first abnormal feature value of each edge pixel on each closed connected domain.
[0029] When the anti-corrosion coating quality of steel components is qualified, the surface of the steel components appears uniform in color and smooth. The gradient angles of the pixels in the surface image of the steel components tend to be consistent due to the uniform distribution of gray levels. When obvious defect areas appear in the acquired surface image of the steel components, the gradient amplitude of the corrosion flaking defect edges is high, while the gradient amplitudes of the pixels on both sides of the defect edge are relatively low, showing a situation where the gradient amplitude is low on both sides of the edge and high at the edge. Similar situations also occur with scratches and cracks. Considering that corrosion flaking defects of steel components are mainly characterized by irregular contours, the gradient angles of the edge pixels corresponding to the defect areas are relatively dispersed. Therefore, this embodiment uses the Sobel operator to calculate the gradient angle and gradient amplitude of each pixel in the surface area of the steel components, calculates the probability that a pixel in the surface area of the steel components is a defect edge point, and then determines whether corrosion flaking defects exist.
[0030] For each edge pixel on each closed connected domain in the surface area of the steel component, taking the u-th edge pixel as an example, a local window of size n×n is set with the u-th pixel as the center, denoted as the local neighborhood, to analyze the possibility of defect edges existing in the local neighborhood of the u-th pixel. The implementer can set the value of n according to the situation. In this embodiment, n=5.
[0031] Based on the above analysis, the first anomaly feature value of each edge pixel in each closed connected domain is determined to characterize the probability that the calculated edge pixel of the connected domain is an edge pixel of corrosion detachment defect, specifically: First, calculate the mean of the difference between the gradient magnitude of each edge pixel on each closed connected region and all pixels in its local neighborhood.
[0032] It should be noted that the difference represents the degree of difference between two variables, which can be calculated using methods such as the difference value, the absolute value of the difference, or the square of the difference. This embodiment calculates the mean of the differences in gradient magnitudes between each edge pixel and all pixels in its local neighborhood on each closed connected component. The expression for the first abnormal feature value of each edge pixel on each closed connected component is: In the formula, Let be the first anomaly feature value of the u-th edge pixel in each closed connected component, sig() be the sigmoid normalization function, and M be the number of pixels in the local neighborhood of the u-th edge pixel in each closed connected component. Let be the mean value of the u-th edge pixel of each closed connected component. Let be the mean value of the a-th pixel within the local neighborhood of the u-th edge pixel of each closed connected component, and Norm() be the normalization function. Let be the gradient angle of the u-th edge pixel in each closed connected region. Let be the gradient angle of the a-th pixel within the local neighborhood of the u-th edge pixel in each closed connected region, and let cos() be the trigonometric cosine function. To ensure that the value is greater than 0 and to avoid a denominator of 0, this embodiment... The implementer can set this according to the actual situation, and this embodiment does not impose any restrictions on it. Specifically, when the a-th pixel in the local neighborhood of the u-th edge pixel is a non-edge pixel, the mean value of the a-th pixel is 0. This is denoted as the first difference.
[0033] It should be understood that when the u-th edge pixel is an edge pixel of an erosion-induced defect, the gradient magnitude of the edge pixel of the erosion-induced defect is higher than that of the pixels in the regions on both sides, resulting in a relatively high first difference value. Furthermore, since the edge of an erosion-induced defect is mainly irregular in shape, the gradient angles are relatively dispersed. In contrast, the gradient angles in the smooth areas outside the defect and the areas inside the defect are relatively close due to the more uniform color change. Therefore, when the u-th edge pixel is a defect edge pixel, there is a significant change in its angle compared to the pixels in its local neighborhood, resulting in a higher calculated first difference value. Typically less than 1, the calculated first abnormal feature value is relatively large. Meanwhile, the gradient directions of other edge pixels and their neighboring pixels are more concentrated, resulting in... As the value approaches 1, the calculated first abnormal feature value becomes relatively small.
[0034] Step S003: Determine the randomness of the peak distribution in the curvature of all edge pixels in each closed connected domain, and determine the edge volatility of each closed connected domain by combining the texture feature differences between each edge pixel and its local neighborhood pixels.
[0035] Peeling or flaking of coatings on steel components is a common coating defect, usually related to the loss of adhesion between the coating and the metal substrate or between old and new coatings. As adhesion gradually weakens, flaking or flake-like detachment occurs on the coating surface, eventually leading to the entire coating peeling off. This phenomenon is called "peeling" or "detachment." When corrosion flaking defects occur in the anti-corrosion coating of steel components, the edges of the flake-like areas may exhibit random, irregular serrated or wavy patterns. Therefore, to determine whether a defect in the anti-corrosion coating is corrosion flaking, taking the c-th closed connected region as an example, the curvature of the edge pixels of the c-th closed connected region is statistically analyzed. The curvature of each edge pixel is calculated, and a random pixel is selected as the starting point. Following a clockwise order, the curvature sequence of the c-th closed connected region can be formed. ,in, This represents the curvature of the first edge pixel of the c-th closed connected component. Let represent the curvature of the t-th edge pixel in the c-th closed connected region, where t represents the total number of edge pixels in the c-th closed connected region.
[0036] This embodiment uses the AMPD (Automatic multiscale-based peak detection) algorithm to obtain each peak in the curvature sequence of the c-th closed connected region, and determines the randomness of the peak distribution in the curvature of all edge pixels of the c-th closed connected region. This is used to characterize the randomness of curvature fluctuations of defective edge pixels. The specific expression is as follows: In the formula, This represents the degree of randomness in the peak distribution of curvature among all edge pixels of the c-th closed connected component. This represents the number of peaks in the curvature sequence of the c-th closed connected region. This represents the number of curvatures between the i-th curvature peak and the (i-1)-th curvature peak in the curvature sequence. This represents the number of curvatures between the i-th curvature peak and the (i+1)-th curvature peak in the curvature sequence. This represents the labeling result of the i-th curvature peak in the curvature sequence. By comparing the number of curvatures between the curvature peak and its preceding and following peaks, when the connected region is an eroded connected region, the number of curvatures between the curvature peaks is relatively random due to its randomly changing irregular sawtooth or wavy characteristics. The calculated value is... Relatively large.
[0037] Meanwhile, since the corrosion peeling defect area is where the coating is peeled off from the surface of the steel component, its edges will be relatively sharp. Brush marks and cracks, on the other hand, do not have the strong local contrast characteristic of the peeling area. Based on the above analysis, taking the c-th closed connected region as an example, the LBP (Local Binary Pattern) value is calculated in a 3×3 window centered on each pixel to determine the texture contrast of each closed connected region. This contrast is used to characterize the grayscale fluctuations and texture changes in the local neighborhood of the edge pixels. The specific expression is as follows: In the formula, Let be the texture contrast of the c-th closed connected region within the surface area of the steel component. This represents the LBP value of the s-th edge pixel in the c-th closed connected component. Let represent the LBP value of the a-th pixel within the local neighborhood centered on the s-th edge pixel in the c-th closed connected component, where N is the number of edge pixels in the c-th closed connected component. The second difference is denoted as the texture contrast, which is the fusion result of the second difference of all edge pixels on the closed connected domain. Fusion means combining multiple variables, which can be achieved by addition, multiplication, addition-multiplication mixing, etc.
[0038] It should be understood that when the c-th closed connected domain is a corrosion peeling defect, since the edge of the peeling area is the separation boundary between the coating and the substrate, the local contrast of the edge is high under illumination, and the LBP value of the edge pixels is high. Meanwhile, the LBP values of other pixels in its local neighborhood are extremely small because there is no large grayscale change, and the calculated texture contrast is relatively large. The edge pixels of brush marks and cracks do not produce obvious light and dark jumps under illumination, and there is no large contrast. Therefore, the LBP values of their edge pixels are relatively small, and the calculated texture contrast is relatively small.
[0039] The edge volatility of the c-th closed connected component is determined by combining the texture contrast of the c-th closed connected component with the randomness, wherein the edge volatility is positively correlated with both the randomness and the texture contrast. In one embodiment, the product of the texture contrast of the c-th closed connected component and the randomness is used as the edge volatility of the c-th closed connected component.
[0040] Step S004: Combine the edge volatility of each closed connected domain with the first abnormal feature value of all edge pixels in each closed connected domain to obtain the second abnormal feature value of each closed connected domain.
[0041] Furthermore, the second abnormal characteristic value of each closed connected region is determined to determine the probability that each closed connected region is a corrosion flaking defect region. The specific expression is as follows: In the formula, This is the second outlier eigenvalue of the c-th closed connected component. This represents the edge volatility of the c-th closed connected component, where Norm() is the normalization function. Let N be the first anomaly feature value of the u-th edge pixel in the c-th closed connected component, and N be the number of edge pixels in the c-th closed connected component. The flowchart for determining the second anomaly feature value of a closed connected component is as follows: Figure 2 As shown.
[0042] It should be understood that the second abnormal feature value combines the edge fluctuation of the closed connected domain with the first abnormal feature value of the edge pixels of the closed connected domain, reflecting the possibility that each closed connected domain is a corrosion flaking defect area. The larger the second abnormal feature value, the higher the possibility that the corresponding closed connected domain is a corrosion flaking defect area, and the worse the anti-corrosion coating quality of the steel component.
[0043] Step S005: Identify the corrosion and flaking areas in the steel component image based on the second abnormal feature value.
[0044] By calculating the second anomaly eigenvalue of each closed connected region on the surface of the steel component in the image acquired by the UAV, and using the Otsu thresholding method to obtain a segmentation threshold for all the acquired second anomaly eigenvalues of the closed connected regions, if there is a closed contour in the steel component image with a second anomaly eigenvalue greater than the segmentation threshold, it is determined that there is a corrosion-scraped area in the steel component image, indicating that the anti-corrosion coating quality of the steel component is unqualified; otherwise, it is determined that there is no corrosion-scraped area in the steel component image, indicating that the anti-corrosion coating quality of the steel component is qualified. The Otsu thresholding method is a well-known existing technology, and the specific process will not be elaborated.
[0045] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments of this specification have been described above. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.
[0046] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.
[0047] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them; modifications to the technical solutions described in the foregoing embodiments, or equivalent substitutions of some of the technical features, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.
Claims
1. An automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs), characterized in that, The method includes the following steps: Based on images of steel components acquired by drones, each closed connected component in the steel component image is obtained; Analyze the gradient difference between each edge pixel and its local neighborhood pixels in each closed connected domain to determine the first abnormal feature value of each edge pixel in each closed connected domain. Determine the randomness of the peak distribution in the curvature of all edge pixels in each closed connected domain, and combine the texture feature differences between each edge pixel and its local neighborhood pixels in each closed connected domain to determine the edge variability of each closed connected domain. By combining the edge volatility of each closed connected domain with the first abnormal feature value of all edge pixels in each closed connected domain, the second abnormal feature value of each closed connected domain is obtained. The corrosion and flaking areas in the steel component image are identified based on the second abnormal feature value.
2. The automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs) as described in claim 1, characterized in that, The determination of the first abnormal feature value includes: The average difference between the gradient magnitude of each edge pixel in each closed connected region and all pixels in its local neighborhood is calculated. Based on the difference between the average difference between each edge pixel in each closed connected region and the edge pixels in its local neighborhood, it is denoted as the first difference. The difference between the gradient angle of each edge pixel in each closed connected region and the pixel in its local neighborhood is used to determine the first abnormal feature value.
3. The automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs) as described in claim 2, characterized in that, The process for determining the first abnormal feature value is as follows: Calculate the difference in gradient angle between each edge pixel on each closed connected domain and each pixel in its local neighborhood, and calculate the square of the cosine of the difference. The first abnormal feature value is negatively correlated with the square of the cosine and positively correlated with the first difference.
4. The automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs) as described in claim 3, characterized in that, The expression for the first abnormal feature value is: In the formula, Let be the first anomaly feature value of the u-th edge pixel in each closed connected component, sig() be the sigmoid normalization function, and M be the number of pixels in the local neighborhood of the u-th edge pixel in each closed connected component. Let be the mean value of the u-th edge pixel of each closed connected component. Let be the mean value of the a-th pixel within the local neighborhood of the u-th edge pixel of each closed connected component, and Norm() be the normalization function. Let be the gradient angle of the u-th edge pixel in each closed connected region. Let be the gradient angle of the a-th pixel within the local neighborhood of the u-th edge pixel in each closed connected region, and let cos() be the trigonometric cosine function. The value is a preset value greater than 0, wherein when the a-th pixel in the local neighborhood of the u-th edge pixel is a non-edge pixel, the mean value of the a-th pixel is 0.
5. The automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs) as described in claim 1, characterized in that, The determination of the randomness of the peak distribution in the curvature of all edge pixels in each closed connected region includes: Arrange the curvature of all edge pixels in each closed connected region in a uniform order, obtain the peak value of the curvature of all edge pixels in each closed connected region after the arrangement, and determine the degree of randomness of the peak distribution of the curvature of all edge pixels in each closed connected region based on the number of curvatures between adjacent peak values.
6. The automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs) as described in claim 5, characterized in that, If the number of curvatures between any peak and its adjacent previous peak is equal to the number of curvatures between any peak and its adjacent next peak, then any peak is marked as 0; otherwise, any peak is marked as 1. The marked values of all peaks of curvatures of all edge pixels of each closed connected domain are summed to determine the degree of randomness of the peak distribution in the curvatures of all edge pixels of each closed connected domain.
7. The automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs) as described in claim 1, characterized in that, Determining the edge volatility of each closed connected component includes: The difference between the LBP values of each edge pixel and its local neighborhood pixels on each closed connected domain is calculated and denoted as the second difference. The fusion result of the second difference of all edge pixels on each closed connected domain is obtained. The edge volatility is positively correlated with the randomness and the fusion result.
8. The automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs) as described in claim 7, characterized in that, The edge volatility of each closed connected domain is the product of the randomness and the fusion result.
9. The automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs) as described in claim 1, characterized in that, The determination of the second abnormal feature value of each closed connected component includes: Calculate the mean of the first abnormal feature values of all edge pixels in each closed connected region, and multiply it by the normalized value of the edge volatility of each closed connected region to obtain the second abnormal feature value of each closed connected region.
10. The automated inspection method for the anti-corrosion coating quality of steel components based on unmanned aerial vehicles (UAVs) as described in claim 1, characterized in that, The process of identifying corrosion and flaking areas in steel component images based on the second abnormal feature value includes: Threshold segmentation is performed on the second abnormal feature value of all closed contours in the steel component image. If there is a closed contour in the steel component image with a second abnormal feature value greater than the segmentation threshold, it is determined that there is a corrosion and flaking area in the steel component image; otherwise, it is determined that there is no corrosion and flaking area in the steel component image.