Metal part bending forming quality evaluation method

Abstract

The invention discloses a metal part bending forming quality evaluation method, which comprises the following steps of: determining the distribution inclusion degree between a gray value and each gray value smaller than the gray value according to the distribution relationship between pixel points corresponding to the gray value in a bending area image and pixel points corresponding to each gray value smaller than the gray value; determining the boundary degree of the gray value according to the distribution inclusion degree between all gray values greater than the gray value and all gray values less than the gray value in the local range of the gray value; correcting the frequency of each gray value according to the boundary degree of each gray value to obtain the corrected frequency of each gray value; enhancing the bent area image according to the correction frequency of each gray value to obtain an enhanced image; and evaluating the bending forming quality of the metal part according to the enhanced image. According to the method, the contrast among different image features in the bending area image is enhanced, and the accuracy of bending forming quality evaluation is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing. More specifically, the present invention relates to a method for evaluating the quality of bent metal parts. Background Art

[0002] During the bending process of metal parts, due to the complex and changeable processing environment, the surface images of the workpieces collected often present the problem of low contrast. Fine creases, cracks and other defects are often hidden in the low-contrast areas, directly affecting the accuracy of subsequent quality inspection. Therefore, it is necessary to enhance the images of the bending areas.

[0003] As a classic image enhancement method, histogram equalization improves the overall contrast by expanding the dynamic range of pixel values and has been widely used in early industrial inspections. However, this method has significant limitations: the histogram equalization algorithm will over-enhance the high-frequency gray-level areas, while the low-frequency gray-level areas may be weakened or even lost. Since fine creases, cracks and other defects on the metal surface often correspond to the low-frequency gray-level value distribution, during the histogram equalization process, these key defect features are easily submerged or completely lost by the background noise, thus seriously affecting the accuracy of the quality evaluation results. Summary of the Invention

[0004] To solve the technical problem that the above-mentioned histogram equalization will cause low-frequency gray-level areas such as creases and cracks to be submerged or completely lost by the background noise, affecting the accuracy of the quality evaluation results, the present invention provides a method for evaluating the quality of bent metal parts, including: For any gray-level value in the image of the bending area, determine the distribution inclusion degree between the gray-level value and the gray-level values less than the gray-level value according to the distribution relationship between the pixel points corresponding to the gray-level value and the pixel points corresponding to the gray-level values less than the gray-level value; Determine the demarcation degree of the gray-level value according to the distribution inclusion degree between all the gray-level values greater than the gray-level value and all the gray-level values less than the gray-level value within the local range of the gray-level value; Correct the frequency of each gray-level value according to the demarcation degree of each gray-level value to obtain the corrected frequency of each gray-level value; Enhance the image of the bending area according to the corrected frequency of each gray-level value to obtain an enhanced image; Evaluate the quality of the bent metal parts according to the enhanced image.

[0005] Preferably, the determination of the distribution inclusion degree between the gray-level value and the gray-level values less than the gray-level value includes: Obtain the minimum circumscribed rectangle corresponding to all the pixel points corresponding to each gray-level value as the minimum circumscribed rectangle of each gray-level value; For any two gray values, obtain the minimum bounding rectangle of these two gray values to obtain the degree of distribution overlap between these two pixel points; use the product of the degree of distribution overlap and the confidence level of the degree of distribution overlap as the degree of distribution inclusion between these two pixel points.

[0006] Preferably, the method for obtaining the degree of distribution overlap is as follows: Obtain the intersection and union between the minimum bounding rectangles of these two gray values, and use the ratio of the number of pixel points contained in the intersection to the number of pixel points contained in the union as the degree of distribution overlap between these two pixel points.

[0007] Preferably, the method for obtaining the confidence level is as follows: Evenly divide the minimum bounding rectangle of the gray value into several grids, count the number of pixel points corresponding to this gray value contained in each grid, obtain the average difference in the number of pixel points corresponding to this gray value contained in any two grids, and perform negative correlation normalization on the average difference to obtain the degree of distribution dispersion of this gray value; For any two pixel points, use the minimum value among the degrees of distribution dispersion of these two pixel points as the confidence level of the degree of distribution overlap between these two pixel points.

[0008] Preferably, determining the degree of demarcation of this gray value includes: Use several gray values after this gray value as the backward reference gray values of this gray value, and use several gray values before this gray value as the forward reference gray values of this gray value; According to the degree of distribution inclusion between the backward reference gray values of this gray value and all the forward reference gray values of this gray value, and the degree of distribution inclusion between the backward reference gray values of this gray value and this gray value and all the backward reference gray values before this backward reference gray value, determine the demarcation probability of this gray value under the backward reference gray value; Use the average value of the demarcation probabilities of this gray value under all the backward reference gray values of this gray value as the degree of demarcation of this gray value.

[0009] Preferably, the method for obtaining the demarcation probability is as follows: For any backward reference gray value of the gray values, obtain the mean of the distribution inclusion degrees between the backward reference gray value and all the forward reference gray values of this gray value as the first mean; obtain the mean of the distribution inclusion degrees between the backward reference gray value and this gray value and all the backward reference gray values before this backward reference gray value as the second mean; take the difference between the second mean and the first mean as the inter-class difference between forward and backward; take the mean of the differences between the distribution inclusion degrees between the backward reference gray value and all the forward reference gray values of this gray value pairwise as the intra-class difference of the forward; take the mean of the differences between the distribution inclusion degrees between the backward reference gray value and this gray value and all the backward reference gray values before this backward reference gray value pairwise as the intra-class difference of the backward; Determine the boundary probability of this gray value under the backward reference gray value according to the inter-class difference between forward and backward, the intra-class difference of the forward, and the intra-class difference of the backward.

[0010] Preferably, the boundary probability satisfies the expression: ; Wherein, represents the boundary probability of the th gray value under the th backward reference gray value; represents the inter-class difference between forward and backward of the th gray value of the th backward reference gray value; represents the intra-class difference of the forward, represents the intra-class difference of the backward; represents the S-shaped curve;

[0011] Preferably, the correction frequency satisfies the expression: ; Wherein, represents the correction frequency of the th gray value, represents the boundary degree of the th gray value; represents the frequency of the th gray value; represents the number of gray values in the bent region image; represents the boundary degree of the th gray value; represents the frequency of the th gray value.

[0012] Preferably, enhancing the image of the bending region according to the correction frequency of each gray value to obtain an enhanced image includes: Taking the gray value as the horizontal axis and the correction frequency of each gray value as the vertical axis, constructing a gray histogram, and performing histogram equalization on the gray histogram to obtain the enhanced image of the bending region image.

[0013] Preferably, evaluating the quality of the metal part bending forming according to the enhanced image includes: Inputting the enhanced image into the trained neural network to output the bending defects of the metal part.

[0014] The beneficial effects of the present invention are as follows: The present invention quantifies the possibility that two gray values belong to the same image feature by the degree of distribution inclusion between the two gray values; based on the degree of distribution inclusion between all gray values greater than the gray value and all gray values less than the gray value within the local range of the gray value, the demarcation degree of the gray value is determined, and the demarcation degree is used to reflect the possibility that the gray value is the demarcation gray value of two image features. Based on the demarcation degree, the frequency of the gray value is corrected, so that the demarcation gray value can be emphasized during enhancement, the contrast between different image features is increased, the over-enhancement of the high-frequency gray area is avoided, and the low-frequency gray value area is not swallowed, the abnormal areas such as cracks and folds can be highlighted, the weak defect features of the bending region are strengthened specifically, and the accuracy of the quality evaluation of the metal part bending forming is improved. Description of the Drawings

[0015] By reading the following detailed description with reference to the drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the drawings, several embodiments of the present invention are shown in an exemplary rather than restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein: Figure 1 is a flowchart schematically showing a method for evaluating the quality of metal part bending forming in the present invention; Figure 2 is a schematic diagram showing the image of the bending region; Figure 3 is a flowchart schematically showing step S2 of a method for evaluating the quality of metal part bending forming in the present invention; Figure 4 is a schematic diagram showing the enhanced image. Detailed Embodiments

[0016] 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 part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the protection scope of the present invention.

[0017] Next, the specific implementation manners of the present invention will be described in detail with reference to the accompanying drawings.

[0018] An embodiment of the present invention discloses a method for evaluating the bending forming quality of metal parts. Referring to Figure 1 , it includes steps S1 to S3: S1. Collect the bending area image of the metal part.

[0019] After the metal part is bent and formed, a uniform illumination system is used to irradiate the bending area of the metal part, and the bending area image of the metal part is collected by a camera. For the convenience of processing, the collected bending area image is a grayscale image. Figure 2 It is a schematic diagram of the bending area image of the metal part.

[0020] S2. Enhance the bending area image.

[0021] Specifically, the flowchart of step S2 is shown in Figure 3 , and it includes steps S201 to S204, specifically: S201. For any gray value in the bending area image, determine the distribution inclusion degree between this gray value and the gray values less than this gray value according to the distribution relationship between the pixel points corresponding to this gray value and the pixel points corresponding to the gray values less than this gray value.

[0022] It should be noted that under Gaussian noise interference, different pixel points of the same structural feature (such as a flat area, a bending area, or a crack, etc.) in the bending area image will show multiple gray value distributions. Although the numerical values of these pixel points with different gray values are different, their spatial distribution patterns still maintain the topological consistency of the feature area. Therefore, the present invention determines the distribution inclusion degree between this gray value and the gray values less than this gray value according to the distribution relationship between the pixel points corresponding to each gray value and the pixel points corresponding to the gray values less than this gray value.

[0023] Specifically, obtain all the pixel points corresponding to each gray value in the bending area image, and obtain the minimum bounding rectangle corresponding to all the pixel points corresponding to each gray value as the minimum bounding rectangle of each gray value.

[0024] The distribution inclusion degree satisfies the expression: ; Where represents the degree of inclusion in the distribution between the th gray value and the th gray value; ; represents the set composed of the pixel points included in the minimum bounding rectangle of the th gray value; represents the set composed of the pixel points included in the minimum bounding rectangle of the th gray value; represents the intersection symbol, represents the union symbol; represents the quantity function; represents the number of pixel points included in the intersection with the set , that is, the number of pixel points included in the overlapping part between the minimum bounding rectangle of the th gray value and the minimum bounding rectangle of the th gray value; represents the number of pixel points included in the union with the set , that is, the number of pixel points included after merging the convex hull region of the th gray value and the minimum bounding rectangle of the th gray value; represents the degree of distribution overlap between the th gray value and the th gray value. When is larger, it indicates that the minimum bounding rectangle of the th gray value and the minimum bounding rectangle of the th gray value overlap more. When is smaller, it indicates that the overlapping degree between the minimum bounding rectangle of the th gray value and the minimum bounding rectangle of the th gray value is smaller. represents the credibility of the degree of distribution overlap between the th gray value and the th gray value. When the credibility is larger, the degree of inclusion in the distribution is closer to the degree of distribution overlap. When the credibility is smaller, the degree of inclusion in the distribution is smaller based on the degree of distribution overlap.

[0025] Furthermore, the credibility satisfies the expression: ; Among them, represents the credibility of the degree of distribution overlap between the th gray value and the th gray value; represents the The degree of dispersion of the pixel points corresponding to a gray value in the minimum bounding rectangle of the th gray value; Indicates the degree of dispersion of the pixel points corresponding to the th gray value in the minimum bounding rectangle of the th gray value. When the degree of dispersion is larger, it indicates that the pixel points corresponding to the gray value show a more dispersed spatial distribution characteristic within its minimum bounding rectangle area. On the contrary, when the degree of dispersion is smaller, it means that the pixel points of the gray value show an obvious spatial aggregation characteristic within the minimum bounding rectangle. When the th gray value and the th gray value are more dispersed in their respective corresponding minimum bounding rectangles, the degree of distribution overlap obtained based on the intersection and union of the minimum bounding rectangles of the th gray value and the th gray value is more credible. On the contrary, when the th gray value is more concentrated in its minimum bounding rectangle, or the th gray value is more concentrated in its minimum bounding rectangle, the degree of distribution overlap is less credible.

[0026] Furthermore, the method for obtaining the degree of dispersion is as follows: The minimum bounding rectangle of the gray value is evenly divided into F grids, and the number of pixel points corresponding to the gray value contained in each grid is counted. Where F is a preset first quantity, which is set by the implementer according to the actual implementation situation. For example , S represents the number of pixel points corresponding to the gray value.

[0027] The degree of dispersion satisfies the expression: ; Among them, represents the degree of dispersion of the pixel points corresponding to the th gray value in the minimum bounding rectangle of the th gray value; represents the number of grids into which the minimum bounding rectangle of the th gray value is divided; represents the th grid of the th gray value and contains the th gray value corresponding to the number of pixel points; represents the th grid of the th gray value and contains the th gray value corresponding to the number of pixel points; represents the The number of pixel points corresponding to a gray value; Denotes the exponential function with the natural constant as the base, and is used for Performing a negative correlation mapping.

[0028] In the formula, Denotes the Average difference in the number of pixel points corresponding to the gray value contained in any two grids of the gray value. When the pixel points corresponding to the gray value are concentrated in one grid, the average difference will be the largest, which is . Therefore, in the present invention, the average difference is multiplied by to achieve the normalization of the average difference. The larger the average difference, the more concentrated the gray value is distributed in its few grids, and the smaller the degree of dispersion. On the contrary, the smaller the average difference, the more grids the

[0029] gray value is distributed in, and the greater the degree of dispersion.

[0030] Specifically, for any gray value in the bent region image, the N gray values after this gray value are used as the backward reference gray values of this gray value, and the N gray values before this gray value are used as the forward reference gray values of this gray value. The backward reference gray values and the forward reference gray values are collectively referred to as reference gray values. When there are not N gray values before this gray value, the forward reference gray value of this gray value is counted according to the actual situation. When there are not N gray values after this gray value, the backward reference gray value of this gray value is counted according to the actual situation, where N is a preset first quantity. In this embodiment, N = 5. In other embodiments, the implementer can set the first quantity according to the actual implementation situation. Then the local range of this gray value is the range composed of its reference gray values and this gray value.

[0031] Furthermore, the degree of demarcation satisfies the expression: ; where Denotes the degree of demarcation of the gray value; Denotes the probability of demarcation of the gray value under the gray value of the gray value; Denotes the number of backward reference gray values of the

[0032] Among them, the th gray value at the th backward reference gray value satisfies the expression for the boundary probability of the th gray value: ; Among them, represents the boundary probability of the th gray value at the th backward reference gray value; represents the between-class difference between the forward and backward directions of the th gray value at the th backward reference gray value, reflecting the degree of distribution inclusion between the th gray value at the th backward reference gray value and all forward reference gray values of the th gray value, and the between-class difference in the degree of distribution inclusion between the th gray value at the th backward reference gray value and all backward reference gray values before the th backward reference gray value (including the th gray value); represents the forward intra-class difference, represents the backward intra-class difference; represents the S-shaped curve, and the expression is ; represents the exponential function with the natural constant as the base. The function is used to map the between-class difference between the forward and backward directions to the range (0, 1). When is negative, is smaller and less than 0.5. When is positive, is larger and greater than 0.5. When is 0, has a value of 0.5. is used to perform a negative correlation mapping on the forward intra-class difference and the backward intra-class difference . When the forward intra-class difference or the backward intra-class difference is larger, the mapping result is smaller. When the forward intra-class difference or the backward intra-class difference is smaller, the mapping result is larger. When the between-class difference between the forward and backward directions is larger, the forward intra-class difference is smaller, and the backward intra-class difference is smaller, the th gray value at the th backward reference gray value​ The greater the boundary probability of the gray value.

[0033] Forward and backward between-class differences Satisfies the expression: ; In the formula, represents the degree of distribution inclusion between the -th gray value and the -th gray value. The -th gray value is the -th backward reference gray value of the -th gray value. represents the average value of the degree of distribution inclusion between the -th backward reference gray value of the -th gray value and all backward reference gray values before the -th backward reference gray value (including the -th gray value); represents the degree of distribution inclusion between the -th gray value and the -th gray value. The -th gray value is the -th forward reference gray value of the -th gray value. represents the average value of the degree of distribution inclusion between the -th backward reference gray value of the -th gray value and all forward reference gray values of the -th gray value.

[0034] Forward within-class differences Satisfies the expression: ; In the formula, represents the degree of distribution inclusion between the -th gray value and the -th gray value; represents the degree of distribution inclusion between the -th gray value and the -th gray value; represents the number of backward reference gray values of the -th gray value. Then the forward within-class difference is the average value of the pairwise differences in the degree of distribution inclusion between the -th backward reference gray value of the -th gray value and all forward reference gray values of the -th gray value.

[0035] Backward within-class differences Satisfy the expression: ; wherein, represents the degree of distribution inclusion between the -th gray value and the -th gray value; represents the degree of distribution inclusion between the -th gray value and the -th gray value; represents the serial number of the backward reference gray value of the -th gray value. Then the backward intra-class difference is the mean of the differences between the pairwise degrees of distribution inclusion between the -th backward reference gray value of the -th gray value and all the backward reference gray values before the -th backward reference gray value (including the -th gray value).

[0036] It should be noted that the degree of distribution inclusion between the gray values in the same image feature in the bent region image is relatively large, while the degree of distribution inclusion between the gray values in different image features is relatively small. If the -th gray value is the boundary gray value between two image features, then the gray values within the local range of the -th gray value that are less than the -th gray value (i.e., the forward reference gray value of the -th gray value) are one kind of image feature, and the gray values within the local range of the -th gray value that are not less than the -th gray value (i.e., the -th gray value and the backward reference gray value of the -th gray value) are another kind of image feature. At this time, the degree of distribution inclusion between the backward reference gray value of the -th gray value and the forward gray value of the -th gray value is relatively small, and the degree of distribution inclusion between the backward reference gray value of the -th gray value and all the backward gray values before the -th gray value and the -th backward reference gray value is relatively large. Therefore, the present invention determines the boundary probability of the -th gray value under the -th backward reference gray value of the -th gray value according to the forward-backward inter-class difference , the forward intra-class difference , and the backward intra-class difference . When the forward-backward inter-class difference The larger it is, it indicates that under the measurement of the th backward reference gray value, all reference gray values within the local range of the th gray value are more likely to belong to two image features. When the forward intra-class difference and the backward intra-class difference are both smaller, it indicates that under the measurement of the th backward reference gray value, the th gray value is more likely to be the boundary gray value of the two image features. Further, when the boundary probability of the th gray value under all backward reference gray values of the th gray value is larger, the boundary degree of the th gray value is greater, and the th gray value is more likely to be the boundary gray value of different image features in the bent region image. The boundary gray value means that several gray values before this gray value are one kind of image feature, and this gray value and several gray values after it are another kind of image feature.

[0037] S203. Modify the frequency of each gray value according to the boundary degree of each gray value to obtain the modified frequency of each gray value.

[0038] It should be noted that the gray values of some image features in the bent region image may be close to those of other image features, resulting in these image features being easily ignored. For example, the fine cracks in the bent region may not be obvious, resulting in missed detection of fine cracks during the detection of the bending forming quality. And the boundary degree of each gray value represents the possibility of each gray value being the boundary gray value of two image features. Therefore, the present invention modifies the frequency of each gray value according to the boundary degree of each gray value, so that the contrast between each image feature in the bent region image after enhancement based on the modified frequency is obvious, thereby improving the accuracy of the detection between the bending forming of metal parts.

[0039] Specifically, the modified frequency of each gray value satisfies the expression: ; Wherein, represents the modified frequency of the th gray value, represents the boundary degree of the th gray value; represents the frequency of the th gray value; represents the number of gray values in the bent region image; represents the boundary degree of the th gray value; represents the frequency of the th gray value; For normalization to ensure that the sum of the correction frequencies of each gray value is 1.

[0040] When the demarcation degree of the th gray value is greater, the th gray value is more likely to be the demarcation gray value between different image features, that is, the th gray value and several subsequent gray values may be an image feature, and the th gray value and several previous gray values may be another image feature. Then when enhancing the image of the bending area, it is necessary to increase the frequency of the th gray value, so that the cumulative frequency of the th gray value in the cumulative distribution histogram of the histogram equalization algorithm increases, so that after histogram equalization, the contrast between the image feature corresponding to the th gray value and the image feature corresponding to the th gray value increases. Therefore, when the demarcation degree of the th gray value is greater, the correction frequency of the th gray value is greater.

[0041] S204. Enhance the image of the bending area according to the correction frequency of each gray value to obtain an enhanced image.

[0042] Taking the gray value as the horizontal axis and the correction frequency of each gray value as the vertical axis, construct a gray histogram, and perform histogram equalization on the gray histogram to obtain an enhanced image of the bending area image. Figure 4 It is a schematic diagram of the enhanced image. It can be seen that Figure 4 there are some horizontal textures, and these horizontal textures are crack defects. Figure 4 The crack defects in Figure 2 are more obvious than those in

[0043] S3. Evaluate the quality of the metal part bending forming according to the enhanced image.

[0044] It should be noted that during the bending forming of the metal part, when the material ductility is insufficient or the bending radius is too small, cracks will occur in the bending area. When the material thickness is too thin or the bending angle is too large, wrinkles will occur in the bending area. Therefore, the present invention identifies cracks and wrinkles in the enhanced image.

[0045] Specifically, in this embodiment, a convolutional neural network is used to identify cracks and wrinkles. The structure of the convolutional neural network is specifically as follows: The input of the convolutional neural network is the enhanced image, the output is the defect recognition result. The dataset of the convolutional neural network is the enhanced images corresponding to the bent region images of metal parts with defects and normal metal parts. The label of the convolutional neural network is the defect recognition result, including no defect, crack, wrinkle, etc. The loss function of the convolutional neural network is the cross-entropy loss.

[0046] Input the enhanced image into the trained convolutional neural network, and output the defect recognition result.

[0047] Thus, the quality evaluation of the metal part bending forming is realized.

Claims

1. A method for evaluating the bending and forming quality of metal parts, characterized in that, Including: For any gray value in the image of the bending area of the metal part, according to the distribution relationship between the pixel points corresponding to the gray value and the pixel points corresponding to each gray value less than the gray value, determine the distribution inclusion degree between the gray value and each gray value less than the gray value; According to the distribution inclusion degree between all gray values greater than the gray value and all gray values less than the gray value within the local range of the gray value, determine the demarcation degree of the gray value; According to the demarcation degree of each gray value, correct the frequency of each gray value to obtain the corrected frequency of each gray value; Enhance the bending area image according to the corrected frequency of each gray value to obtain an enhanced image; Evaluate the bending forming quality of the metal part according to the enhanced image.

2. The quality evaluation method for bending and forming of metal parts according to claim 1, wherein The determination of the distribution inclusion degree between the gray value and each gray value less than the gray value includes: Obtain the minimum bounding rectangle corresponding to all pixel points corresponding to each gray value as the minimum bounding rectangle of each gray value; For any two gray values, obtain the distribution coincidence degree between the minimum bounding rectangles of these two gray values and between these two pixel points; take the product of the distribution coincidence degree and the credibility of the distribution coincidence degree as the distribution inclusion degree between these two pixel points.

3. The quality evaluation method for bending and forming of metal parts according to claim 2, characterized in that, The method for obtaining the distribution coincidence degree is: Obtain the intersection and union between the minimum bounding rectangles of these two gray values, and take the ratio of the number of pixel points contained in the intersection to the number of pixel points contained in the union as the distribution coincidence degree between these two pixel points.

4. A method for evaluating the bending forming quality of metal parts according to claim 2, characterized in that, The method for obtaining the credibility is: Evenly divide the minimum bounding rectangle of the gray value into several grids, count the number of pixel points corresponding to the gray value contained in each grid, obtain the average difference in the number of pixel points corresponding to the gray value contained in any two grids, and perform negative correlation normalization on the average difference to obtain the distribution dispersion degree of the gray value; For any two pixel points, take the minimum value among the distribution dispersion degrees of these two pixel points as the credibility of the distribution coincidence degree between these two pixel points.

5. A method for evaluating the bending and forming quality of a metal part according to claim 1, characterized in that The determination of the demarcation degree of the gray value includes: Take several gray values after the gray value as the backward reference gray values of the gray value, and take several gray values before the gray value as the forward reference gray values of the gray value; According to the distribution inclusion degree between the backward reference gray values of the gray value and all the forward reference gray values of the gray value, and the distribution inclusion degree between the backward reference gray values of the gray value and the gray value and all the backward reference gray values before the backward reference gray value, determine the demarcation probability of the gray value under the backward reference gray value; Take the average value of the demarcation probabilities of the gray value under all the backward reference gray values of the gray value as the demarcation degree of the gray value.

6. A method for evaluating the bending and forming quality of metal parts according to claim 5, characterized in that, The method for obtaining the demarcation probability is: For any backward reference gray value of the gray value, obtain the mean of the distribution inclusion degrees between the backward reference gray value and all forward reference gray values of this gray value as the first mean; obtain the mean of the distribution inclusion degrees between the backward reference gray value and this gray value and all backward reference gray values before this backward reference gray value as the second mean; take the difference between the second mean and the first mean as the inter-class difference between forward and backward; take the mean of the differences between the distribution inclusion degrees between the backward reference gray value and all forward reference gray values of this gray value pairwise as the intra-class difference of the forward; take the mean of the differences between the distribution inclusion degrees between the backward reference gray value and this gray value and all backward reference gray values before this backward reference gray value pairwise as the intra-class difference of the backward; Determine the boundary probability of this gray value under the backward reference gray value according to the inter-class difference between forward and backward, the intra-class difference of the forward, and the intra-class difference of the backward.

7. A method for evaluating the bending and forming quality of metal parts according to claim 6, characterized in that, The boundary probability satisfies the expression: ； Among them, represents the th gray value of the th backward reference gray value under the th gray value of the boundary probability; represents the th gray value of the th backward reference gray value of the forward and backward between-class differences; represents the forward within-class difference, represents the backward within-class difference; represents the S-shaped curve; represents the exponential function with the natural constant as the base.

8. A method for evaluating the bending and forming quality of a metal part according to claim 1, characterized in that, The correction frequency satisfies the expression: ； Among them, represents the correction frequency of the th gray value, represents the demarcation degree of the th gray value; represents the th gray value frequency; represents the number of gray values in the image of the bending area; represents the demarcation degree of the th gray value; represents the th gray value frequency.

9. A method for evaluating the bending and forming quality of metal parts according to claim 1, characterized in that, Enhancing the image of the bending region according to the correction frequency of each gray value to obtain an enhanced image, including: Taking the gray value as the horizontal axis and the correction frequency of each gray value as the vertical axis, constructing a gray histogram, and performing histogram equalization on the gray histogram to obtain the enhanced image of the bending region image.

10. A method for evaluating the bending and forming quality of metal parts according to claim 1, characterized in that, Evaluating the quality of the metal part bending forming according to the enhanced image, including: Inputting the enhanced image into the trained neural network and outputting the bending defect of the metal part.

Images