Buffer braid damage area detection method based on image segmentation

Abstract

The invention belongs to the technical field of image processing, and particularly relates to a buffer braid damage area detection method based on image segmentation, and the method comprises the steps: synchronously collecting braid depth images through an upper depth camera and a lower depth camera, and obtaining a target image and a relative target image; obtaining the regularity of the target image according to the texture distribution rule of the target image, and calculating the texture regularity of the braid through the regularity mean value of all depth image pairs obtained by the regularity mean value of the two target images on the braid; according to a difference value between the target image regularity and the braid texture regularity and a proportion of the difference value in the braid texture regularity, calculating an abnormal possibility of the target image; in combination with the abnormal possibility of the target image and the probability that all suspected areas and relative suspected areas in the target image and the relative target image belong to defect areas, the defect degree of the braid is obtained and marked; according to the invention, the detection and positioning precision of the damaged area of the buffer braid is improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more particularly to a method for detecting damaged areas of a buffer webbing based on image segmentation. Background Art

[0002] When electricians work at height, buffer webbing is a critical line of defense for maintaining life safety, making its quality inspection crucial. In the event of a fall, the buffer webbing must utilize a scientifically designed structure and high-quality materials to instantly and effectively absorb the impact and minimize harm. To ensure the safety of power workers, the buffer webbing on protective gear must be regularly inspected for safety. Internal damage to the buffer webbing may result in bulging (disordered accumulation of internal fibers) or localized thinning (fracture and wear of internal fibers). These defects are not readily apparent on the webbing surface. If undetected, these defects can easily lead to stress concentration during a fall, causing the bulge or thinning to occur first, leading to fracture or structural collapse.

[0003] To detect defects in materials like cushioning webbing, machine vision is currently commonly used to scan the webbing surface. Related technologies, such as Chinese patent publication CN112200790B, disclose methods, devices, and media for fabric defect detection. This disclosure utilizes a convolutional neural network to separately acquire spatial and detail feature data and fuse them to determine fabric defects, improving detection accuracy.

[0004] However, the patented fabric defect detection methods, equipment and media have problems when facing complex damage conditions of buffer webbing, such as local thinning and bulging caused by internal fiber breakage. Due to the lack of consideration of the physical structure and three-dimensional deformation of the webbing, the detection accuracy is low, which poses a hidden danger to power workers working at height. Summary of the Invention

[0005] In order to solve the problem that the existing technology is difficult to detect the local thinning and bulging of the ribbon caused by the breakage of the internal fibers of the ribbon, the present invention provides a buffer ribbon damaged area detection method based on image segmentation, comprising: The method uses two upper and lower depth cameras to synchronously capture webbing depth images to obtain several depth image pairs, which constitute a depth image set. The images contained in any depth image pair are recorded as target images and relative target images. Based on the depth distribution, the target image is divided to obtain several feature areas and suspected areas. According to the distribution law of the feature areas of the target image, the feature area distance is obtained. The feature area distance is used to judge the depth similarity of all feature areas on the target image and calculate the regularity of the target image. The regularity of the webbing texture is calculated by averaging the regularity means of all pairs of target and relative target images. The difference between the regularity of the target image and the regularity of the webbing texture is compared to calculate the probability of target image anomaly. The depth extreme point is located in the suspected area, a perpendicular line to the extreme point is established, and the perpendicular plane of the perpendicular line is moved to generate a scanning plane. The number of pixels in each scanning plane is counted to obtain a pixel number sequence. Combined with the probability of target image anomaly, the probability of the suspected area being a defect is calculated. The probability of the relatively suspected area being a defect is located and calculated. The degree of webbing defect is obtained by combining the probability of all suspected and relatively suspected areas in all depth image pairs being defective areas, and the damaged area is marked.

[0006] Existing technologies have difficulty detecting defects such as local thinning and bulging caused by internal fiber breakage when inspecting buffer webbing, which can easily reduce the accuracy of webbing inspection. The present invention uses a depth camera to capture double-sided depth images of the webbing. By calculating multiple parameters such as target image regularity, webbing texture regularity, the likelihood of target image anomalies, and the probability that suspected areas are defects, it comprehensively determines the degree of webbing defects and determines the location of these damages, greatly improving the accuracy and reliability of buffer webbing damage detection. This invention is highly targeted, specifically designed for buffer webbing, fully considering its importance in high-altitude operations and common damage forms such as webbing wear and thinning. Using a depth camera to capture depth images, it can detect internal damage, better meeting the inspection needs of buffer webbing.

[0007] Preferably, the calculating target image regularity includes: Calculate the feature area distance; the target image regularity satisfies the following expression: ; Where, Indicates the regularity of the target image; Indicates the number of all feature region sets in the target image; Indicates the number of feature regions in the i-th feature region set; represents the depth mean of the jth feature region in the i-th feature region set, Represents the mean depth of the target image; It represents the depth mean of the feature region that is separated from the jth feature region by s feature regions in the i-th feature region set; Represents a very small positive number, ensuring that the denominator is not 0; represents the absolute value function; Represents the normalization function.

[0008] Preferably, obtaining the characteristic region distance includes: Points smaller than the depth mean in the target image are extracted and grown to form feature regions. The central pixel point of the central feature region of the target image is used to generate a 0°~180° rotated straight line with the central pixel point as the midpoint. The straight line containing the most feature regions is screened out as the target line. The feature regions on the straight line are the target set, and the target set period is calculated based on the number of target set regions and the length of the target line. The target line is translated multiple times until the image boundary to obtain multiple target set periods and the average of all target set periods of the target image is calculated to represent the target image period, which is recorded as the feature region distance.

[0009] Preferably, the regularity of the webbing texture includes: Obtain the regularity of the relative target image in the depth image pair where the target image is located; calculate the mean of the regularity of the target image and the relative target image to obtain the texture regularity of the local area of ​​the ribbon; then calculate the texture regularity of the ribbon through the local regularity of all depth image pairs.

[0010] The present invention utilizes the ability to first clarify the rules of the overall texture of the ribbon, such as the intervals between the textures and the periodicity of changes in depth, to provide a reference standard for subsequent judgment of where damage may exist, making it easier to find abnormal areas that deviate from the template.

[0011] Preferably, the probability of abnormality of the target image satisfies the following expression: ; Where, Indicates the probability of abnormality of the target image; Indicates the regularity of the ribbon texture; Indicates the regularity of the target image; Represents a very small positive number, ensuring that the denominator is not 0; represents the absolute value function; Represents the normalization function.

[0012] The present invention can quantify the degree of damage to the regularity of local texture when the webbing is damaged, and determine whether a certain area may have abnormalities. It is equivalent to conducting a preliminary screening first, marking suspected problematic areas, narrowing the scope of subsequent key inspections, and improving inspection efficiency.

[0013] Preferably, dividing the suspected area includes: A region growing algorithm is performed on pixel points in the target image whose depth is greater than or equal to the depth mean to generate a first suspected region representing a suspected thinning region; a region growing algorithm is performed on pixel points whose depth is less than the depth mean to generate a second suspected region representing a suspected bulge region; the first suspected region and the second suspected region are collectively defined as a suspected region.

[0014] Preferably, the calculating the probability that the suspected area is a defect includes: Get the pixel count sequence; the probability that the suspected area is a defect satisfies the following expression: ; Where, 、 、 The horizontal plane generated by the corresponding position on the baseline after the moving point p moves for the zth, vth, and v+1th times, and the number of pixels swept by the horizontal plane on the suspected area; represents the total number of times the moving point p moves on the baseline; Represents the difference between the depth of the k-th pixel in the suspected area and the mean depth of the relative target image; Indicates the probability of abnormality of the target image where the suspected area is located; Represents the normalization function.

[0015] The present invention scans and analyzes the extreme depth points within the suspected area and records the changes in the number of pixels scanned across the horizontal plane. This is equivalent to performing a CT scan on the suspected area. It can capture the shape, size and other characteristics of the area in detail, and further analyze the changes in the number of pixels, providing a solid basis for determining whether it is a real defect.

[0016] Preferably, the step of obtaining a pixel number sequence includes: The depth extreme point p is located in the suspected area, and a baseline perpendicular to the camera plane is established with p as the base point. A horizontal plane is generated by moving in the direction of the vertical plane of the baseline with a step size d, and the number of pixels swept by the horizontal plane of p in the suspected area after each movement is recorded. After three initial movements, if the number of pixels is greater than 0 for three consecutive times, the point moves in the original direction until it equals 0; if it is less than or equal to 0 for three consecutive times, the point moves in the opposite direction. The number of all movements of the moving point p and the number of pixels swept by the horizontal plane of the moving point p in the suspected area are extracted to obtain the pixel number sequence of the suspected area.

[0017] Preferably, the locating and calculating the probability that the relatively suspected area is a defect includes: The suspected area in the target image relative to the target image is located and recorded as the relative suspected area, and the probability that the relative suspected area is a defect is calculated.

[0018] Preferably, the degree of webbing defect satisfies the following expression: ; Where, Indicates the degree of webbing defects; Q indicates the number of depth image pairs; Indicates the number of defective areas on the target image; represents the wth depth image pair on the target image The probability that a suspected area is a defect; represents the wth depth image pair relative to the target image. The probability that a relatively suspected area is a defect; Represents the normalization function.

[0019] The present invention can combine the conditions of suspected areas in the images taken by the upper and lower cameras, verify each other, and finally comprehensively evaluate the degree of defects of the entire webbing, ensuring more accurate detection results, reducing misjudgments or missed judgments, and finally clearly marking the damaged area for easy subsequent processing.

[0020] Compared to the existing patent CN112200790B, this invention can accurately detect damaged areas on the cushioning webbing, such as bulges and localized thinning, and precisely locate these damaged areas, greatly improving the accuracy and reliability of cushioning webbing damage detection. For workers working at heights, cushioning webbing is a critical safety feature, and this method can promptly identify potential hazards and prevent accidents caused by webbing damage. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flow chart schematically illustrating a method for detecting damaged areas of a buffer webbing based on image segmentation in the present invention; Figure 2 FIG. 1 is a schematic diagram schematically showing the camera position. FIG. DETAILED DESCRIPTION

[0022] The embodiment of the present invention discloses a method for detecting damaged areas of a buffer webbing based on image segmentation, referring to Figure 1 , including steps S1 to S4: S1: Use the upper and lower depth cameras to synchronously capture the webbing depth image to obtain the target image and relative target image.

[0023] It should be noted that damage to the buffer webbing may be hidden internally, such as localized thinning caused by fiber breakage or surface anomalies such as bulges. Ordinary visual inspection has difficulty capturing depth differences. Therefore, a depth camera is used to simultaneously capture images from the top and bottom sides to determine the regularity of the buffer webbing's texture. The difference between the regularity of the target image and the regularity of the overall webbing texture is compared, the probability of an anomaly in the target image is calculated, and suspected areas (bulges or thinning areas) are divided according to the relationship between the depth value and the mean. The depth extreme points in the suspected areas are scanned and analyzed. This method can not only obtain the surface shape but also reflect internal structural changes through depth values, ensuring comprehensive coverage of possible damage areas.

[0024] It's important to note that the buffer webbing worn by power workers working at height requires regular inspection. While existing image processing technology can identify surface flaws and wear in used high-altitude safety belts through image enhancement, edge detection, and texture analysis, it has limitations in detecting internal damage. This step uses two depth cameras, one above the other, to capture a pair of local depth images of the buffer webbing. The depth image sets are then continuously moved with equal steps.

[0025] Specifically, a used buffer webbing to be tested is recovered; two depth cameras are placed above and below the webbing respectively, so that the optical axis of the camera is perpendicular to the plane of the webbing and the field of view completely overlaps, such as Figure 2 The figure is a schematic diagram of the camera positions; a synchronization mechanism is triggered to ensure that the two cameras simultaneously capture images of the ribbon at the same position, eliminating position deviations caused by time differences; the upper and lower camera images are aligned and denoised; a depth image pair is obtained, which includes two upper and lower depth images, recorded as the target image and the relative target image respectively; the buffer ribbon is continuously moved in equal steps along the direction of the ribbon extension to obtain several depth image pairs, forming a depth image set. It should be noted that each pixel value in the depth image represents the vertical distance from the point to the camera, usually in millimeters. The smaller the pixel depth value, the closer the vertical distance from the point to the camera; the larger the pixel depth value, the farther the vertical distance from the point to the camera.

[0026] At this point, a set of depth images of the buffer webbing to be detected is obtained.

[0027] S2: According to the texture distribution law of the target image, the target image regularity is obtained, and the texture regularity of the ribbon is calculated by taking the regularity mean of all depth image pairs obtained by the regularity mean of the two target images on the ribbon.

[0028] It should be noted that the buffer webbing is composed of interwoven warp and weft threads, forming a regular texture. This texture appears as evenly spaced light and dark areas in the target image, with the light areas representing ridges and the dark areas representing valleys. The depth values ​​of the plain weave exhibit a periodic "ridge-valley-ridge" pattern. This periodicity is reflected in the fact that the depth values ​​of the initial feature area in the target image, after moving the corresponding period in any direction along the alternating light and dark areas, are similar to those of the initial feature area. Therefore, the present invention utilizes this characteristic to derive a method for representing the regularity of the buffer webbing texture.

[0029] Specifically, a target image in a depth image pair is selected from a depth image set, and the mean depth value of all pixels in the target image is calculated, which is recorded as the depth mean. In the target image, pixels with depth values ​​less than the depth mean are extracted to obtain multiple feature pixels. Region growing is performed on these feature pixels to obtain multiple feature regions. A minimum bounding rectangle is generated for all feature regions, and half the diagonal length of the minimum bounding rectangle of each feature region is calculated and recorded as the core distance. It should be noted that by performing region growing on these feature pixels, adjacent feature points of the same type can be aggregated to form a feature region, which more clearly reflects the overall distribution of the texture. The minimum bounding rectangle is generated and the core distance is calculated. The core distance is the maximum radius of the feature region and is used to define the spatial range of the region. It can quantify the size range of each feature region and provide a standard for subsequently determining whether the feature regions belong to the same rotation line, ensuring that the selected target line can accurately reflect the true distribution direction and pattern of the ribbon texture.

[0030] Preferably, the center pixel point of the central feature area of ​​the target image is used as the origin, and multiple straight lines rotating around the origin are generated, and the straight lines rotate every 1°, covering the range of 0°~180°; the vertical distance between each rotating straight line and the center pixel point of the feature area is calculated, and if the distance is less than or equal to the core distance, the feature area is determined to belong to the current straight line, and the area is recorded as the target area; the number of target areas corresponding to each straight line is counted; the average value of the target areas of all straight lines is calculated, and the straight line with the highest number of target areas and its corresponding direction are screened out, and the straight line is recorded as the target straight line; all target areas on the target straight line are recorded as a target set, and the average distance between adjacent target areas on the target straight line is calculated by comparing the length of the target straight line with the number of target areas on the target straight line and recorded as the characteristic target set period. It should be noted that the texture of the warp and weft interweaving of the buffer webbing has significant periodicity, which is reflected in the regular spacing of the feature areas in the same direction and the regular distribution in the vertical direction. By generating a 0°~180° rotated straight line and screening the target line with the most feature areas, the direction of the texture target line can be accurately captured, and the interval pattern of the direction can be quantified by calculating the feature target set period; by translating the target line in the vertical direction and taking the average of multiple set periods, the texture characteristics in different directions can be comprehensively reflected, thereby more comprehensively extracting the feature area distance.

[0031] Preferably, the distance between the center pixel points of any two adjacent feature areas perpendicular to the extension direction of the target straight line is calculated, and this distance is set as the translation step size of the target straight line on both sides perpendicular to the extension direction of the target straight line, and the target straight line is translated to both sides along the vertical direction until it reaches the image boundary and stops, obtaining multiple translated straight lines and corresponding multiple sets; the average of all set periods is calculated to represent the average distance of all feature areas of the target image, which is recorded as the feature area distance.

[0032] It's important to note that while the texture of the buffer webbing exhibits a periodic distribution along the target line, it also exhibits regular patterns perpendicular to the line. Using the distance between the centers of adjacent feature regions in the vertical direction as the translation step ensures that the translated line still captures the valid feature regions while also matching the vertical distribution density of the webbing texture.

[0033] Preferably, the average depth of the feature regions in each set in the target image is obtained, and the similarity of the average depth of the feature regions separated from the feature regions by multiple feature region distances is calculated to represent the regularity of the target image. The regularity of the target image satisfies the following expression: ; Where, Indicates the regularity of the target image; Indicates the number of all feature region sets in the target image; Indicates the number of feature regions in the i-th feature region set; represents the depth mean of the jth feature region in the i-th feature region set, Represents the mean depth of the target image; It represents the depth mean of the feature region that is separated from the jth feature region by s feature regions in the i-th feature region set; Represents a very small positive number, ensuring that the denominator is not 0; represents the absolute value function; Represents the normalization function.

[0034] Where, It is used to measure the degree of deviation between the depth values ​​of the jth feature region and the feature region separated by s feature regions from the jth feature region in the i-th feature region set; Indicates the square of the deviation of the depth mean of the jth feature region in the i-th feature region set from the depth mean of the target image; The closer it is to 1, the more consistent the depth deviations between the jth feature region and the feature region separated by s feature regions from the jth feature region in the i-th feature region set are, that is, the greater the similarity of the depth values ​​of the two regions. When the depth deviations of the two regions are in opposite directions or have large differences in magnitude, the fractional value approaches 0, that is, the less similar the depth values ​​of the two regions are. Represents the mean similarity of the depth values ​​between the jth feature region in the i-th feature region set and all feature regions in the i-th feature region set, which is recorded as the similarity of the i-th feature region; Represents the mean similarity of all feature regions in the i-th feature region set; It represents the average of the similarity means of all feature regions in the set of all feature regions in the target image, which is used as the regularity of the target image. It should be noted that this expression quantifies this regularity by calculating the similarity of the deviation degree between the mean depth of the feature region separated by s feature regions and the mean depth of the target image. The closer the similarity is to 1, the stronger the regularity. The regularity of the target image obtained by averaging and normalizing the similarity means of all feature region sets reflects the neatness of the ribbon texture.

[0035] Preferably, the regularity of the relative target image in the depth image pair where the target image is located is obtained; the mean of the regularity of the target image and the relative target image is calculated, and the texture regularity of the local area of ​​the ribbon corresponding to the target image and the relative target image is obtained through the mean of the regularity of the target image and the relative target image; the mean of the regularity of the target image and the relative target image in all depth image pairs in the depth image set is calculated, and recorded as the mean of the regularity of the depth image pair; the average of the regularity means of all depth image pairs is calculated, and the average is recorded as the texture regularity of the ribbon. It should be noted that the texture regularity of the buffer ribbon as a continuous whole should be consistent in the local area and the whole. By obtaining the regularity of the relative target image and calculating the mean together with the regularity of the target image, the texture regularity of the local ribbon area can be comprehensively reflected, avoiding the errors that may exist in a single image. Further calculating the average of the regularity means of all depth image pairs as the texture regularity of the ribbon can fully reflect the overall texture regularity of the entire ribbon.

[0036] At this point, the regularity of the ribbon texture is obtained.

[0037] S3: Calculate the probability of target image abnormality based on the difference between the target image regularity and the ribbon texture regularity, and the proportion of the difference in the ribbon texture regularity.

[0038] It should be noted that the manufacturing process of buffer webbing dictates that a normal product possesses regular structural features, such as texture periodicity, thickness uniformity, and consistent fiber orientation. However, defects such as bulging or localized thinning can disrupt the regularity of these features. Specifically, localized bulging or thinning can cause abrupt changes in the contour curve in the 2D image, disrupting the regular structural features of the buffer webbing and reducing the regularity of the target image. This step determines whether a localized area of ​​the webbing exhibits surface bulging or localized thinning based on whether it conforms to the regular texture features of the buffer webbing. The degree of abnormality in the localized area is recorded as the probability of target image abnormality.

[0039] Specifically, the regularity of the ribbon texture and the regularity of any target image in the depth image set are extracted, and the probability of the target image being abnormal is calculated. The probability of the target image being abnormal satisfies the following expression: ; Where, Indicates the probability of abnormality of the target image; Indicates the regularity of the ribbon texture; Indicates the regularity of the target image; Represents a very small positive number, ensuring that the denominator is not 0; represents the absolute value function; Represents the normalization function.

[0040] Where, Indicates the difference between the regularity of the ribbon texture and the regularity of the target image. The larger the value, the less the regularity of the target image is close to the regularity of the ribbon texture, and the smaller the regularity of the target image, the greater the possibility of the target image being abnormal. The larger the value, the smaller the regularity of the target image and the greater the possibility of abnormality of the target image. It should be noted that this formula accurately captures the abnormal characteristics of the local area by quantifying the degree of deviation between the regularity of the target image and the regularity of the ribbon texture.

[0041] S4: Combining the probability of abnormality of the target image and the probability that all suspected areas and relatively suspected areas in the target image and the relative target image belong to defective areas, the degree of webbing defect is obtained and marked.

[0042] It should be noted that this step combines the probability of abnormality in the target image with the defect probability of suspected areas and relatively suspected areas, enabling a comprehensive assessment from local to global perspective. The relatively suspected areas in the upper and lower images confirm each other, reducing misjudgments caused by single-angle shooting. The final normalized calculation of the webbing defect level directly reflects the severity of damage to the entire webbing, accurately marking damaged areas and meeting the inspection requirements of the buffer webbing as a continuous whole.

[0043] Specifically, the mean depth value in the target image is extracted, and all pixels with depth values ​​greater than or equal to the mean depth value are subjected to region growing to generate a first suspected region, which represents the suspected thinning area. All pixels with depth values ​​less than the mean depth value are subjected to region growing to generate a second suspected region, which represents the suspected bulge area. Both the first suspected region and the second suspected region are referred to as suspected regions. It should be noted that under normal conditions, the thickness distribution of the buffer webbing is relatively uniform, and the depth values ​​of the pixels in the depth image should show regular fluctuations around the mean depth value. When the webbing is locally thinned, the depth value of this area will be greater than or equal to the mean depth value; when a bulge defect exists, the depth value of this area will be less than the mean depth value.

[0044] Preferably, any suspected area is selected, and the pixel point p with the extreme depth value is located, and the depth of point p is recorded as . Take the three-dimensional position of point p as the base point, the upward direction of point p perpendicular to the camera as the height reduction direction, and the straight line perpendicular to the camera starting from point p is recorded as the baseline of p; suppose there is a moving point on the baseline of p, which moves multiple times at a uniform speed starting from point p along the baseline, with a moving step of d. The plane perpendicular to the baseline where the moving point is located is recorded as the horizontal plane of the moving point. When the moving point moves once, the number of pixels swept by the horizontal plane of the moving point in the suspected area is recorded. Move three times continuously. If the number of pixels is continuously greater than 0, continue to move the moving point in this direction. If the number of pixels is continuously less than or equal to 0, continue to move the moving point in the opposite direction. The stopping condition of the movement is that the number of pixels is 0; extract all the movement times of the moving point p, and the number of pixels swept by the horizontal plane corresponding to each movement of the moving point in the suspected area, and obtain the pixel number sequence of the suspected area; analyze the change of the number of pixels in the pixel number sequence of the suspected area, and obtain the probability that the suspected area is a defect. The probability that the suspected area is a defect satisfies the following expression: ; Where g represents the probability that the suspected area is a defect; It represents the number of pixels swept over the suspected area by the horizontal plane generated at the corresponding position on the baseline after the moving point p moves for the vth time, and e represents the number of pixels swept over the suspected area by the horizontal plane; Indicates the number of pixels swept across the suspected area by the horizontal plane generated by the corresponding position on the baseline after the moving point p moves for the v+1th time; The sum of the number of times the moving point p moves on the baseline; The number of pixels corresponding to the z-th movement of the moving point p; Represents the difference between the depth of the k-th pixel in the suspected area and the mean depth of the relative target image; Indicates the probability of abnormality of the target image; represents the normalization function; Where, Indicates the total number of pixels corresponding to the z-th movement of the moving point p. The larger the value, the larger the area of ​​the suspected area. It represents the difference between the number of pixels swept by the horizontal plane generated at the corresponding position on the baseline after the moving point p moves for the v+1th time and the number of pixels swept by the horizontal plane generated at the corresponding position on the baseline after the moving point p moves for the vth time; It represents the sum of the changes in the pixels swept across the suspected area by the horizontal plane generated when the moving point p moves n times on the baseline; It represents the average change in the number of pixels swept across the suspected area by the horizontal plane generated by the moving point p when it moves n times on the baseline. The larger the value, the greater the degree to which the suspected area belongs to the defect area. Represents the sum of the differences between the depth of each pixel in the suspected area and the mean depth of the target image. A value greater than 0 indicates that the suspected area is a bulge area, and a value less than 0 indicates that the suspected area is a thinning area. The larger it is, the greater the probability that the suspected area is a defect. It should be noted that the damaged area of ​​the buffer webbing has significant three-dimensional morphological differences in depth features, and it is difficult to express the degree of the defect only through a plane image. This step locates the depth extreme point of the suspected area and moves the horizontal plane along the baseline for scanning. The sequence of the number of recorded pixel points can intuitively reflect the range and morphological changes of the defect in the depth direction. After moving three times in a row, the direction is adjusted according to the number of pixels to ensure complete coverage of the defect area and avoid missing the defect area. The expression of the probability that the suspected area is a defect combines the area size, morphological changes, depth deviation and the possibility of abnormality of the target image, realizing a method of integrating the physical characteristics of the defect and the abnormal trend of the webbing texture, so that the calculation result accurately reflects the possibility that the suspected area is a real defect.

[0045] Preferably, the target image is located relative to the suspected area in the target image and recorded as the relative suspected area. The probability that the relative suspected area is a defect is calculated using the same method as the probability that the suspected area is a defect. It should be noted that the buffer webbing damage has corresponding regional features in the front and back depth images, and can be verified in both the front and back directions by locating the relative suspected area.

[0046] Preferably, the credibility of the suspected area being a defect is analyzed by the relationship between the probability that the suspected area in the target image is a defect and the probability that the relative suspected area is a defect; the credibility of all suspected areas in the target image being defects is calculated as the defect credibility of the depth image pair where the target image is located; the defect credibility of all depth image pairs in the depth image set is calculated as the degree of webbing defect. The degree of webbing defect satisfies the following expression: ; Where, Indicates the degree of webbing defects; Q indicates the number of depth image pairs; Indicates the number of defective areas on the target image; represents the wth depth image pair on the target image The probability that a suspected area is a defect; represents the wth depth image pair relative to the target image. The probability that a relatively suspected area is a defect; Represents the normalization function.

[0047] Where, represents the wth depth image pair The suspected area and The sum of the probabilities of the relatively suspected areas belonging to defects is recorded as Defect credibility of each suspected area; Represents all the w-th depth image pair The sum of the defect credibility of the suspected areas is recorded as the defect credibility of the w-th depth image pair; The sum of the defect credibility of Q w-th depth image pairs is recorded as the ribbon defect credibility; finally, Indicates the degree of webbing defects. It should be noted that the buffer webbing, as a continuous entity, may have defects distributed across different regions, and a single depth image pair cannot fully reflect the overall damage condition. By combining the defect probabilities of suspected areas in the target image and the relative target image, the defect credibility of a single depth image pair is calculated, which can reflect the damage degree of that local area. Further calculation and normalization of the defect credibility of all depth image pairs yields the webbing defect degree, which comprehensively covers the damage condition of the entire webbing.

[0048] Preferably, a first threshold is preset, if If the value is greater than the first threshold, it indicates that there is a damaged area in the webbing, and all damaged areas of this buffer webbing are marked. It should be noted that the first threshold is set by the implementer based on the actual implementation situation. For example, the first threshold can be set to 0.8. When the calculated webbing defect degree F is greater than 0.8, it means that the overall damage to the webbing has exceeded the safety range and there is a risk of affecting safety in use. At this time, marking all damaged areas can not only clearly point out the specific problem location, facilitating subsequent repair or scrapping assessment, but also ensure that the test results match the safety requirements in actual applications, avoiding safety accidents caused by unclear judgment of the defect degree.

[0049] At this point, a buffer webbing damaged area detection method based on image segmentation has been completed.

[0050] Although this specification has shown and described several embodiments of the present invention, it is obvious to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and substitutions without departing from the idea and spirit of the present invention.

Claims

1. A buffer webbing damaged area detection method based on image segmentation, characterized in that: include: Use the upper and lower depth cameras to synchronously collect the ribbon depth image to obtain several depth image pairs to form a depth image set. The image contained in any depth image pair is recorded as the target image and the relative target image; Based on the depth distribution, the target image is divided to obtain several feature areas and suspected areas; according to the distribution law of the feature areas of the target image, the feature area distance is obtained; the feature area distance is used to judge the depth similarity of all feature areas on the target image, and the target image regularity is calculated; the regularity of the ribbon texture is calculated by averaging the regularity means of all targets and the relative target image; the difference between the target image regularity and the ribbon texture regularity is compared to calculate the probability of abnormality of the target image; the depth extreme point is located in the suspected area, a perpendicular line to the extreme point is established and the perpendicular plane of the perpendicular line is moved to generate a scanning plane, the number of pixels in each scanning plane is counted, a pixel number sequence is obtained, and the probability of the suspected area being a defect is calculated in combination with the probability of abnormality of the target image; the probability of the relative suspected area being a defect is located and calculated; The degree of webbing defect is obtained by combining the probability that all suspected and relatively suspected areas in all depth image pairs belong to defect areas, and the damaged areas are marked.

2. The method for detecting damaged areas of a buffer webbing based on image segmentation according to claim 1, characterized in that: The calculating target image regularity includes: Obtain the feature area distance; the target image regularity satisfies the following expression: ； Where, Indicates the regularity of the target image; Indicates the number of all feature region sets in the target image; Indicates the number of feature regions in the i-th feature region set; represents the depth mean of the jth feature region in the i-th feature region set, Represents the mean depth of the target image; It represents the depth mean of the feature region that is separated from the jth feature region by s feature regions in the i-th feature region set; Represents a very small positive number, ensuring that the denominator is not 0; represents the absolute value function; Represents the normalization function.

3. The method for detecting damaged areas of a buffer webbing based on image segmentation according to claim 2, characterized in that: The obtaining of the characteristic area distance includes: Points smaller than the depth mean in the target image are extracted and grown to form feature regions. The central pixel point of the central feature region of the target image is used to generate a 0°~180° rotated straight line with the central pixel point as the midpoint. The straight line containing the most feature regions is screened out as the target line. The feature regions on the straight line are the target set, and the target set period is calculated based on the number of target set regions and the length of the target line. The target line is translated multiple times until the image boundary to obtain multiple target set periods and the average of all target set periods of the target image is calculated to represent the target image period, which is recorded as the feature region distance.

4. The method for detecting damaged areas of a buffer webbing based on image segmentation according to claim 1, characterized in that: The calculation of the ribbon texture regularity includes: The regularity of the relative target image in the depth image pair where the target image is located is obtained; the mean regularity of the target image and the relative target image is calculated to obtain the texture regularity of the local area of ​​the ribbon, and then the texture regularity of the ribbon is calculated by taking the mean of the local regularity of all depth image pairs.

5. The method for detecting damaged areas of a buffer webbing based on image segmentation according to claim 1, characterized in that: The probability of the target image being abnormal satisfies the following expression: ； Where, Indicates the probability of abnormality of the target image; Indicates the regularity of the ribbon texture; Indicates the regularity of the target image; Represents a very small positive number, ensuring that the denominator is not 0; represents the absolute value function; Represents the normalization function.

6. The method for detecting damaged areas of a buffer webbing based on image segmentation according to claim 1, characterized in that: The obtaining of the suspected area includes: A region growing algorithm is performed on pixel points in the target image whose depth is greater than or equal to the depth mean to generate a first suspected region representing a suspected thinning region; a region growing algorithm is performed on pixel points whose depth is less than the depth mean to generate a second suspected region representing a suspected bulge region; the first suspected region and the second suspected region are collectively defined as a suspected region.

7. The method for detecting damaged areas of a buffer webbing based on image segmentation according to claim 1, characterized in that: The calculating the probability that the suspected area is a defect includes: Get the pixel number sequence; The probability that the suspected area is a defect satisfies the following expression: ； Where, 、 、 The horizontal plane generated by the corresponding position on the baseline after the moving point p moves for the zth, vth, and v+1th times, and the number of pixels swept by the horizontal plane on the suspected area; The total number of times the moving point p moves on the baseline; Represents the difference between the depth of the k-th pixel in the suspected area and the mean depth of the relative target image; Indicates the probability of abnormality of the target image where the suspected area is located; Represents the normalization function.

8. The method for detecting damaged areas of a buffer webbing based on image segmentation according to claim 7, characterized in that: The step of obtaining a pixel number sequence includes: The depth extreme point p is located in the suspected area, and a baseline perpendicular to the camera plane is established with p as the base point. A horizontal plane is generated by moving in the direction of the vertical plane of the baseline with a step size d, and the number of pixels swept by the horizontal plane of p in the suspected area after each movement is recorded. After three initial movements, if the number of pixels is greater than 0 for three consecutive times, the point moves in the original direction until it equals 0; if it is less than or equal to 0 for three consecutive times, the point moves in the opposite direction. The number of all movements of the moving point p and the number of pixels swept by the horizontal plane of the moving point p in the suspected area are extracted to obtain the pixel number sequence of the suspected area.

9. The method for detecting damaged areas of a buffer webbing based on image segmentation according to claim 1, characterized in that: The positioning and calculation of the probability that the suspected area is a defect include: The suspected area of ​​the target image is located in the relative target image and recorded as the relative suspected area. The probability that the relative suspected area is a defect is calculated.

10. The method for detecting damaged areas of buffer webbing based on image segmentation according to claim 1, characterized in that: The degree of webbing defect satisfies the following expression: ； Where, Indicates the degree of webbing defects; Q indicates the number of depth image pairs; Indicates the number of defective areas on the target image; represents the wth depth image pair on the target image The probability that a suspected area is a defect; represents the wth depth image pair relative to the target image. The probability that a relatively suspected area is a defect; Represents the normalization function.

Images