Bumper fixing support punch forming quality detection method

Abstract

The invention relates to the technical field of image data processing, in particular to a bumper fixing support punch forming quality detection method, which comprises the following steps: acquiring a surface image of a bumper fixing support; determining the suspected noise degree of each pixel point; determining the position sensitivity degree of each pixel point; determining the window demand degree of each pixel point; determining the self-adaptive window side length of each pixel point; and based on the self-adaptive window side length, filtering the surface image by using a guide filtering algorithm so as to realize punch forming quality detection of the bumper fixing bracket. By dynamically adjusting the size of the filtering window of each target pixel point, the precision of the guiding filtering algorithm is improved, the defect of excessive smoothness or insufficient noise suppression caused by a fixed window is avoided, it is ensured that details of burr defects are reserved, interference between noise and real defects is reduced while denoising is conducted, and the accuracy of the guiding filtering algorithm is improved. And the accuracy of subsequent quality detection is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing. More specifically, the present invention relates to a method for detecting the stamping forming quality of a bumper fixing bracket. Background Art

[0002] In the automotive manufacturing industry, especially in the production process of body parts, as one of the important body parts, the production quality of the bumper fixing bracket plays a crucial role in the safety and appearance of the whole vehicle. In the stamping forming process of the bumper fixing bracket, burr defects are inevitable common problems, and these burrs will not only affect the functionality of the bracket, but may also have an impact on subsequent assembly and use. However, the noise in the image of the bumper fixing bracket will interfere with the recognition of potential burr defects, resulting in a decrease in the detection accuracy of subsequent defect detection. As a boundary-preserving filtering method, the guided filtering algorithm can effectively smooth the image while retaining detail and boundary information, improving the accuracy and reliability of burr defect detection for the bumper fixing bracket and reducing false alarms and missed detections.

[0003] However, when using the guided filtering algorithm to denoise the surface image of the bumper fixing bracket, for each pixel point in the surface image, a window of a fixed size will be selected, and the size of the window will control the smoothing degree of the guided filtering algorithm for this pixel point during filtering; when the window is larger, the filtering will be smoother but the boundary details may be relatively blurred; when the window is smaller, the boundary details can be better retained, but the denoising effect will be relatively poor. Therefore, for the pixel points potentially located at the burr defect on the bumper fixing bracket, if a larger window is selected, it will cause the details of the burr defect to be lost in the filtered image, resulting in the inability to accurately detect all potential burr defects in the image, leading to missed detections and the like in subsequent defect detection; if a smaller window is selected, it will lead to insufficient noise suppression ability in the image, so that the noise is not effectively filtered, and the remaining noise will be confused with the characteristics of the burr defect, making it difficult to accurately distinguish the true burr defect from the noise interference in the subsequent defect detection process, resulting in missed alarms (such as missing the detection of the true burr defect due to noise masking) or false alarms (such as misjudging the noise as a burr), thereby affecting the accuracy and reliability of the quality detection of the stamping forming of the bumper fixing bracket. Summary of the Invention

[0004] To solve the problem that when the surface image of the bumper fixing bracket is denoised using the guided filtering algorithm, a window of a fixed size is used. When the window is large, it is easy to lose the burr details, resulting in missed detections. When the window is small, the denoising is not thorough, and the noise is confused with the defect features, causing missed reports or false alarms, thus affecting the accuracy and reliability of the final quality inspection results. The present invention proposes a method for detecting the stamping forming quality of a bumper fixing bracket, which includes the following steps: Obtain the surface image of the bumper fixing bracket; Denote any pixel point in the surface image as the target pixel point, obtain the defective pixel points among the neighboring pixel points of the target pixel point, and determine the suspected noise degree of the target pixel point according to the sum of the gray value differences between the target pixel point and its neighboring pixel points, and the number of the defective pixel points; Determine the position sensitivity degree of the area where the target pixel point is located according to the standard deviation of the gradient amplitudes of all the pixel points with gradient changes in the area where the target pixel point is located, and the distance between the target pixel point and the pixel point located in the basic direction of the target pixel point and at the boundary of the surface image; Determine the window requirement degree of the target pixel point according to the suspected noise degree, the position sensitivity degree, and a preset value adjustment coefficient; Weight the initial window side length in the guided filtering algorithm according to the window requirement degree to obtain the adaptive window side length of the target pixel point; Based on the adaptive window side length, use the guided filtering algorithm to filter the surface image, and then use a deep learning algorithm to detect the defects in the filtered surface image of the bumper fixing bracket to complete the detection of the stamping forming quality of the bumper fixing bracket.

[0005] The present invention dynamically adjusts the window side length of the guided filtering algorithm according to the suspected noise degree and the position sensitivity degree of the target pixel point, realizes the adaptive configuration of the filter window size, enables the guided filtering algorithm to effectively suppress noise, improve the denoising effect, and maintain the detail information at the burr defect boundary, avoiding detail blurring; Dynamically adjusting the window size can accurately distinguish real burr defects from image noise, reduce false alarms caused by noise interference, and at the same time avoid the situation of missed detections due to the loss of defect details caused by a large window, thereby improving the accuracy and reliability of defect detection; Using adaptive guided filtering combined with deep learning for defect detection, fully retaining defect features and removing noise, greatly improves the recognition ability of surface defects in the stamping forming of bumper fixing brackets and the overall quality of detection.

[0006] Further, the obtaining method of the surface image is: perform graying processing on the collected original image of the bumper fixing bracket, and then use a semantic segmentation model to extract the original image of the bumper fixing bracket after graying processing to obtain the surface image of the bumper fixing bracket.

[0007] Further, the suspected noise degree: ; wherein, is the suspected noise degree of the th pixel point, is the sum of the differences between the gray values of the th pixel point and all its neighboring pixel points, is the number of defective pixel points among the neighboring pixel points of the th pixel point, is the difference between the gray value of the th pixel point and the maximum value among the gray values of all pixel points in the surface image, is the difference between the gray value of the th pixel point and the minimum value among the gray values of all pixel points in the surface image, is a hyperparameter, is the standard normalization function, is the minimum value function.

[0008] By introducing the combined calculation of the sum of the differences between the pixel point and the neighboring gray values, the number of defective pixels in the neighborhood, and the differences between the gray value of this pixel point and the maximum and minimum gray values of the whole image, the present invention can more comprehensively and accurately reflect the noise characteristics of this pixel point, scientifically measure the suspected noise degree; by using the normalization processing and the introduction of the minimum value function, the instability of the calculation caused by the zero or extremely small gray value difference is avoided, making the evaluation of the suspected noise degree more accurate, effectively distinguishing real defects from noise interference, and improving the reliability of noise detection; Furthermore, the acquisition method of the defective pixel points is: the neighboring pixel points with a gray value difference greater than a preset defect threshold from the target pixel point among the neighboring pixel points of the target pixel point are recorded as defective pixel points.

[0009] Furthermore, the basic directions include horizontally left, horizontally right, vertically up, and vertically down.

[0010] Furthermore, the position sensitivity degree satisfies: ; wherein, is the position sensitivity degree of the region to which the th pixel point belongs, is the standard deviation of the gradient amplitudes of all pixel points with gradient changes within the region to which the th pixel point belongs, are respectively the distances between the th pixel point and the pixel points located in the horizontally left, horizontally right, vertically up, and vertically down directions of the target pixel point and on the boundary of the surface image, is the natural exponential function, is the minimum value function.

[0011] The present invention quantifies the position sensitivity of a target pixel in the spatial structure by combining the standard deviation of the gradient magnitude to which the target pixel belongs and the minimum value of the distance from the pixel to the boundaries of multiple directional images, and can accurately reflect the importance of the pixel at the image boundary and gradient change; the exponential decay function is used to combine the gradient change and the boundary distance, highlighting the regions near the image boundary and with large gradient changes, effectively improving the attention to regions with rich details or complex structures, and preventing key regions from being blurred during filtering.

[0012] Further, the distance is the Euclidean distance.

[0013] Further, the window requirement degree satisfies: ; where is the window requirement degree of the th pixel, is the suspected noise degree of the th pixel, is the position sensitivity of the th pixel, is a preset value adjustment coefficient, is the standard normalization function.

[0014] The window requirement degree of the present invention reflects the comprehensive demand of the pixel for the size of the filtering window by combining the suspected noise degree and the position sensitivity, realizing a reasonable balance of multi-dimensional information; the normalization function and the value adjustment coefficient are used to normalize and fine-tune the result, making the size of the filtering window adaptively change, which can not only increase the window to achieve effective smoothing in regions with high noise and low sensitivity to details, but also reasonably reduce the window in boundary-sensitive regions to protect the image details.

[0015] Further, the side length of the adaptive window satisfies: ; where is the side length of the adaptive window of the th pixel, is the initial side length of the window in the guided filtering algorithm, is the window requirement degree of the th pixel, is the combination function used to round up non-odd values to the nearest odd number.

[0016] In the present invention, by rounding up the calculated adaptive window side length to the nearest odd number, it is ensured that the window size meets the requirements of the guided filtering algorithm for an odd window side length, which is helpful for the symmetry and algorithm stability in the filtering process; by combining the initial window side length with the window demand degree, the filtering window size corresponding to each pixel point is dynamically adjusted, making the filtering operation more flexible and personalized, and meeting the different noise suppression and detail protection requirements of different pixel points; the adaptive adjustment of the window side length can effectively avoid the problems of excessive smoothing or detail loss caused by a fixed window, enhance the protection of burr defect details in the filtering process, and improve the noise suppression effect at the same time.

[0017] Further, the deep learning algorithm adopts a convolutional neural network model.

[0018] The present invention has the following beneficial effects: (1) By dynamically adjusting the filtering window size of each target pixel point, the present invention can balance the requirements of denoising and boundary retention according to the suspected noise degree and position sensitivity of the image, thereby improving the accuracy of the guided filtering algorithm and avoiding the defects of excessive smoothing or insufficient noise suppression caused by a fixed window; a larger window can smooth the image, but it is easy to blur details, especially small defects such as burrs, and a smaller window helps to retain details, but the denoising effect is relatively poor. The present invention optimizes through an adaptive window to ensure that the details of burr defects are retained, reduce the interference between noise and real defects while denoising, and improve the accuracy of burr defect detection.

[0019] (2) By comprehensively considering the suspected noise degree, gradient change, and position sensitivity of the target pixel point, the present invention adaptively adjusts the filtering window, effectively enhancing the denoising effect of the image, while ensuring that the boundary features and burr defects are not weakened, avoiding false alarms and missed detections, and improving the accuracy of subsequent defect detection; by using a deep learning algorithm to combine with the high-quality image after guided filtering for defect detection, the accuracy of the detection result can be effectively improved, reducing missed detections and false alarms caused by noise masking or incorrect determination, and enhancing the reliability and stability of stamping forming quality detection.

[0020] (3) Through the optimized image processing technology, the present invention can provide a more accurate input image for subsequent defect detection, improve the efficiency of quality detection, help reduce manual intervention, ensure the quality stability of automotive parts, and improve production efficiency. Description of the Drawings

[0021] Figure 1 is a step flow chart of a method for detecting the stamping forming quality of a bumper fixing bracket according to an embodiment of the present invention. Detailed Embodiments

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below. The described embodiments are part of the embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0023] The specific implementation manners of the present invention will be described in detail below with reference to the accompanying drawings.

[0024] Please refer to Figure 1 , which shows a flowchart of the steps of a method for detecting the stamping forming quality of a bumper fixing bracket provided by an embodiment of the present invention. The method includes the following steps: S01: Obtain the surface image of the bumper fixing bracket.

[0025] It should be noted that the present invention uses a high-definition camera to collect the original image of the bumper fixing bracket.

[0026] Specifically, the manner of obtaining the surface image is as follows: Perform grayscale processing on the collected original image of the bumper fixing bracket, and then use a semantic segmentation model to extract the grayscale-processed original image of the bumper fixing bracket to obtain the surface image of the bumper fixing bracket.

[0027] Implementers can select a semantic segmentation model according to specific implementation situations. For example, the FCN semantic segmentation model.

[0028] S02: Determine the suspected noise level of each pixel point.

[0029] It should be noted that the more extreme the relative size of the grayscale value of each pixel point in the image, the greater the possibility that it is affected by noise, and the greater the corresponding suspected noise level. However, since most of the bumper fixing brackets are black, the grayscale values of their pixel points are also extremely low, and the noise pixel points are usually randomly sporadic; therefore, if the difference in grayscale values between a pixel point and its neighboring pixel points is greater, it indicates that the possibility of being affected by noise is greater, and the corresponding suspected noise level is greater. However, there may be potential burr defects on the surface of the bumper fixing bracket, and the burr defects are usually in the shape of slender spikes, so some pixel points in the area of potential burr defects may also have a large difference in grayscale values from their neighboring pixel points while having relatively extreme grayscale values. However, the pixel points in the area of potential burr defects usually have pixel points with close grayscale values in a specific neighboring direction. Then, in order to exclude the influence of the pixel points in the area of potential burr defects on the quantification of the suspected noise level of a pixel point, that is, in order to distinguish noise pixel points from pixel points in the area of potential burr defects, if the number of neighboring pixel points of a pixel point whose grayscale difference from it exceeds the defect threshold is larger, it indicates that the possibility of this pixel point being affected by noise is greater, the possibility of belonging to the area of potential burr defects is smaller, and the corresponding suspected noise level is greater.

[0030] Denote any pixel point in the surface image as the target pixel point, obtain the defective pixel points among the neighboring pixel points of the target pixel point, and determine the suspected noise level of the target pixel point according to the sum of the differences in grayscale values between the target pixel point and the neighboring pixel points of the target pixel point, and the number of the defective pixel points.

[0031] The implementer can set the number of neighboring pixel points according to the specific implementation situation. For example, 8 neighboring pixel points; if a certain pixel point is located at the boundary of the surface image, the neighboring pixel points beyond the image boundary of this pixel point are obtained by mirror reflection filling.

[0032] Specifically, the way to obtain the defective pixel points is as follows: Denote the neighboring pixel points among the neighboring pixel points of the target pixel point whose grayscale difference from the target pixel point is greater than the preset defect threshold as defective pixel points.

[0033] The implementer can set the defect threshold according to the specific implementation situation. For example, 10.

[0034] Specifically, the suspected noise level: ; In the formula, is the suspected noise level of the th pixel point, is the sum of the differences in grayscale values between the th pixel point and all its neighboring pixel points, is the The number of defective pixels among the neighboring pixels of a pixel is the difference between the gray value of the -th pixel and the maximum value among the gray values of all pixels in the surface image, is the difference between the gray value of the -th pixel and the minimum value among the gray values of all pixels in the surface image, is a hyperparameter, is the standard normalization function, is the minimum value function.

[0035] Implementers can set the hyperparameter according to the specific implementation situation. For example, 0.001. The existence of the hyperparameter is to prevent from causing the formula to be meaningless.

[0036] Among them, the smaller it is, the more extreme the relative size of the gray value of the -th pixel in the surface image, indicating that the possibility of it being affected by noise is greater, and the corresponding suspected noise level is higher; vice versa. the larger it is, the greater the credibility that the relative size of the gray value of the -th pixel in the surface image is more extreme, indicating that the possibility of it being affected by noise is greater, and the corresponding suspected noise level is higher; vice versa. the larger it is, the greater the credibility that the -th pixel is more likely to be affected by noise, the smaller the possibility that it belongs to the potential burr defect area, and the higher the corresponding suspected noise level; vice versa.

[0037] S03: Determine the position sensitivity of each pixel.

[0038] It should be noted that since there may be burr defects in the bumper fixing bracket, in order to prevent the guided filtering algorithm from over-smoothing the originally potential burr defects in the surface image, thus affecting the final defect detection result. Therefore, in this step, the texture features of the region to which each pixel belongs will continue to be analyzed, and the position sensitivity of the region to which each pixel belongs will be calculated. This index is used to reflect the possibility of burr defects at the position of the pixel. Then, based on the position sensitivity and the suspected noise level corresponding to each pixel, the window requirement level of each pixel will be calculated, which is used for calculating the window side length of each pixel in the subsequent steps; then, when analyzing and calculating the position sensitivity of each pixel, since burr defects usually exist at the outer contour edge of the bracket, so for the pixels in the outer contour edge region, more details should be retained as much as possible. For these pixels, a smaller window is required for filtering, so their window requirement level will also be smaller.

[0039] Determine the position sensitivity of the region to which the target pixel belongs according to the standard deviation of the gradient amplitudes of all pixels with gradient changes in the region to which the target pixel belongs, and the distance between the target pixel and the pixel located on the basic direction of the target pixel and at the boundary of the surface image.

[0040] Implementers can set the size of the region according to the specific implementation situation. For example, the region where 7×7 pixels centered on the target pixel are located; if a certain pixel is located at or near the boundary of the surface image, the part of the region to which the pixel belongs that exceeds the image boundary is filled by mirror reflection.

[0041] Specifically, the basic directions include horizontally left, horizontally right, vertically up, and vertically down.

[0042] Specifically, the distance is the Euclidean distance.

[0043] Specifically, the position sensitivity satisfies: ; In the formula, is the position sensitivity of the region to which the th pixel belongs, the th pixel is the standard deviation of the gradient amplitudes of all pixels with gradient changes in the region to which the pixel belongs (if there are no pixels with gradient changes in the region to which a certain pixel belongs, then let , that is, do not consider the existence of this index), are respectively the distances between the th pixel and the pixels located on the horizontally left, horizontally right, vertically up, and vertically down directions of the target pixel and at the boundary of the surface image, is the natural exponential function, is the minimum value function.

[0044] Among them, the smaller it is, the greater the unity of the gradient magnitudes of the pixels with gradient changes in the region to which the th pixel belongs. Then, the greater the possibility that the region to which this pixel belongs is in the outer contour boundary region, and the greater the possibility that there are potential burr defects in the region to which this pixel belongs, and the greater the corresponding position sensitivity; vice versa. the smaller it is, the greater the credibility that the region to which the th pixel belongs is in the outer contour boundary region, the greater the possibility that there are potential burr defects in the region to which this pixel belongs, and the greater the corresponding position sensitivity; vice versa.

[0045] S04: Determine the window requirement degree of each pixel.

[0046] It should be noted that the window requirement degree is a key indicator for balancing the image denoising effect and the retention of burr defect details. In order to overcome the problems of missed reports and false reports caused by the fixed window size of traditional guided filtering. Therefore, in this step, a dynamic window requirement evaluation mechanism is constructed by integrating the suspected noise degree and the position sensitivity of the pixel. The suspected noise degree is used to quantify the possibility of the pixel being interfered by noise, while the position sensitivity reflects the probability of the existence of burr defects in the region where the pixel is located. By combining the two indicators, a large window can be used in the high-noise region to enhance the denoising effect, and the window can be adaptively reduced in the potential burr defect region to retain details, so as to achieve the optimal balance between denoising and feature protection, and provide an accurate basis for the dynamic adjustment of the guided filtering window size in the subsequent steps.

[0047] According to the suspected noise degree and the position sensitivity, and a preset value adjustment coefficient, determine the window requirement degree of the target pixel.

[0048] Specifically, the window requirement degree satisfies: ; In the formula, is the window requirement degree of the th pixel, is the suspected noise degree of the th pixel, is the position sensitivity of the th pixel, is the preset value adjustment coefficient, is the standard normalization function.

[0049] Implementers can set the value adjustment coefficient according to the specific implementation situation. For example, Among them, the larger it is, it indicates that the th pixel point has a greater possibility of being affected by noise. Then, a larger window should be set for this pixel point during filtering to perform a greater degree of smoothing. Therefore, the window requirement degree of this pixel point is greater. the larger it is, it indicates that the th pixel point has a greater possibility of being located in the outer contour edge area. Since there is a greater possibility of potential burr defects in the area where this pixel point is located, a smaller window is required for this area during filtering to retain more detailed information. Therefore, the window requirement degree of this pixel point will also be smaller.

[0050] S05: Determine the adaptive window side length of each pixel point.

[0051] It should be noted that the greater the window requirement degree of each pixel point, it indicates that this pixel point has a greater possibility of being affected by noise and a greater possibility of being located in an area with simple texture and not easy to generate burrs. Then, a larger window should be set for this pixel point to perform a greater degree of filtering to improve the denoising effect; the smaller the window requirement degree of each pixel point, it indicates that this pixel point has a smaller possibility of being affected by noise and a greater possibility of being located in an area with complex texture and easy to generate burrs. Then, a smaller window should be set for this pixel point to retain details as much as possible.

[0052] According to the window requirement degree, the initial window side length in the guided filtering algorithm is weighted to obtain the adaptive window side length of the target pixel point.

[0053] Specifically, the adaptive window side length satisfies: ; In the formula, is the adaptive window side length of the th pixel point, is the initial window side length in the guided filtering algorithm, is the window requirement degree of the th pixel point, is a combination function used to round up a non-odd value to the nearest odd number.

[0054] Implementers can set the initial window side length in the guided filtering algorithm according to the specific implementation situation. For example, 7.

[0055] S06: Based on the adaptive window side length, use the guided filtering algorithm to filter the surface image to achieve the stamping forming quality detection of the bumper fixing bracket.

[0056] It should be noted that after the adaptive calculation of the window size for each pixel point in the surface image is completed, in this step, the optimized guided filtering algorithm will be used to filter the surface image of the bumper fixing bracket, and then based on the filtered surface image of the bumper fixing bracket, potential burr defects on the bracket surface will be detected.

[0057] Based on the adaptive window side length, use the guided filtering algorithm to filter the surface image, and then through the deep learning algorithm, detect defects in the filtered surface image of the bumper fixing bracket to complete the quality detection of the stamping forming of the bumper fixing bracket.

[0058] Implementers can set the regularization parameter in the guided filtering algorithm according to the specific implementation situation. For example, 0.05.

[0059] Specifically, the deep learning algorithm uses a convolutional neural network model (Convolutional Neural Network, CNN).

[0060] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for detecting the stamping forming quality of a bumper fixing bracket, characterized in that Including: Obtain the surface image of the bumper fixing bracket; Denote any pixel point in the surface image as the target pixel point, obtain the defective pixel points among the neighboring pixel points of the target pixel point, and determine the suspected noise degree of the target pixel point according to the sum of the gray value differences between the target pixel point and its neighboring pixel points, as well as the number of the defective pixel points; Determine the position sensitivity of the area where the target pixel point is located according to the standard deviation of the gradient amplitudes of all pixel points with gradient changes in the area where the target pixel point is located, and the distance between the target pixel point and the pixel point located on the basic direction of the target pixel point and at the boundary of the surface image; Determine the window requirement degree of the target pixel point according to the suspected noise degree, the position sensitivity, and a preset value adjustment coefficient; Weight the initial window side length in the guided filtering algorithm according to the window requirement degree to obtain the adaptive window side length of the target pixel point; Based on the adaptive window side length, use the guided filtering algorithm to filter the surface image, and then use the deep learning algorithm to detect the defects in the surface image of the filtered bumper fixing bracket, so as to complete the stamping forming quality detection of the bumper fixing bracket.

2. A method for detecting the stamping quality of a bumper fixing bracket according to claim 1, characterized in that, The acquisition method of the surface image is: Perform gray-scale processing on the original image of the collected bumper fixing bracket, and then use the semantic segmentation model to extract the original image of the gray-scale processed bumper fixing bracket to obtain the surface image of the bumper fixing bracket.

3. A method for detecting the stamping quality of a bumper fixing bracket according to claim 1, characterized in that The suspected noise degree: ； Wherein, is the suspected noise level of the th pixel point, is the sum of the differences between the gray values of the th pixel point and all its neighboring pixel points, is the number of defective pixel points among the neighboring pixel points of the th pixel point, is the difference between the gray value of the th pixel point and the maximum value among the gray values of all pixel points in the surface image, is the difference between the gray value of the th pixel point and the minimum value among the gray values of all pixel points in the surface image, is a hyperparameter, is the standard normalization function, is the minimum value function.

4. A method for detecting the stamping quality of a bumper fixing bracket according to claim 1 or 3, characterized in that, The acquisition method of the defective pixel points is: Denote the neighboring pixel points with a gray difference greater than a preset defect threshold among the neighboring pixel points of the target pixel point as defective pixel points.

5. A method for detecting the stamping quality of a bumper fixing bracket according to claim 1, characterized in that, The basic directions include horizontally to the left, horizontally to the right, vertically upward, and vertically downward.

6. A method for detecting the stamping quality of a bumper fixing bracket according to claim 5, characterized in that, The position sensitivity satisfies: ； In the formula, is the position sensitivity of the region to which the th pixel belongs, is the standard deviation of the gradient magnitudes of all the pixels with gradient changes in the region to which the th pixel belongs, are the distances between the th pixel and the pixels located at the boundaries of the surface image in the horizontal left, horizontal right, vertical up, and vertical down directions of the target pixel, is the natural exponential function, is the minimum value function.

7. A method for detecting the stamping quality of a bumper fixing bracket according to claim 1 or 6, characterized in that, The distance is the Euclidean distance.

8. A method for detecting the stamping quality of a bumper fixing bracket according to claim 1, characterized in that The window requirement degree satisfies: ； Wherein, is the window requirement degree of the th pixel point, is the suspected noise degree of the th pixel point, is the position sensitivity degree of the th pixel point, is a preset value adjustment coefficient, is a standard normalization function.

9. A method for detecting the stamping quality of a bumper fixing bracket according to claim 1, characterized in that, The adaptive window side length satisfies: ； In the formula, is the side length of the adaptive window of the th pixel point, is the initial window side length in the guided filtering algorithm, is the window requirement degree of the th pixel point, is a combination function used to round up a non-odd value to the nearest odd number.

10. A method for detecting the stamping quality of a bumper fixing bracket according to claim 1, characterized in that, The deep learning algorithm uses a convolutional neural network model.

Images