A method for intelligent evaluation of tarpaulin quality
By calculating the local eigenvalues and feature distribution of the waterproof fabric coating and combining them with a decision tree model, the problem of minor imperfections being submerged in the background texture and affecting the accuracy of the assessment was solved, thus improving the accuracy of the waterproof fabric quality assessment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- OGGE (BEIJING) TECH CO LTD
- Filing Date
- 2025-11-12
- Publication Date
- 2026-04-17
AI Technical Summary
In existing technologies, when assessing the quality of waterproof fabrics, minor imperfections are easily obscured by the background texture, leading to insufficient accuracy in the assessment results.
Crack characteristic values are calculated using local directional change rate and local grayscale difference. Small hole characteristic values are determined by combining LBP distribution entropy and edge closure degree. Bubble characteristic values are obtained by identifying the bubble center point and grayscale difference. The quality of waterproof fabric is comprehensively evaluated using a decision tree model.
It improves the accuracy of identifying minor defects in the waterproof fabric coating, thereby enhancing the accuracy of waterproof fabric quality assessment.
Smart Images
Figure CN121481982B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of image processing technology, specifically to an intelligent evaluation method for the quality of waterproof fabric. Background Technology
[0002] Waterproof fabric is a strong and durable textile made of waterproof material with a waterproof coating applied to one side. To ensure the waterproofing effect and durability of projects using waterproof fabric, its quality needs to be assessed. Generally, waterproof fabric coatings are prone to defects such as cracks, pinholes, and bubbles. Existing technologies typically use texture extraction algorithms, such as gray-level co-occurrence matrix (GLCM), to extract texture features from images of the waterproof fabric coating surface, and then identify these defects based on the extracted texture features.
[0003] However, cracks, pinholes, and bubbles in the waterproof fabric coating are all minor defects that are easily obscured by the background texture during evaluation, leading to missed or false detections of these minor defects and affecting the accuracy of the waterproof fabric quality assessment results. Summary of the Invention
[0004] This application provides an intelligent quality assessment method for waterproof fabrics to solve the problem that minor defects are easily obscured by background textures, affecting the accuracy of the quality assessment results. The specific technical solution adopted is as follows:
[0005] One embodiment of this application provides a method for intelligent quality assessment of waterproof fabric, the method comprising the following steps:
[0006] Acquire images of the waterproof fabric coating on the surface to be evaluated;
[0007] Denote any pixel in the waterproof fabric coating image as the target pixel, establish a local analysis window for the target pixel, and determine the local direction change rate and local grayscale difference of the target pixel based on the gradient difference of all edge pixels and the pixel value difference of all pixels in the local analysis window of the target pixel. Calculate the crack feature value of the waterproof fabric coating image based on the local direction change rate and local grayscale difference of all pixels in the waterproof fabric coating image.
[0008] Based on the difference in LBP values of all edge pixels along the same edge in the waterproof fabric coating image, the LBP distribution entropy of the same edge is determined. Based on the difference in the number of pixels contained in the region divided by the edge and the outer circle of the divided region in the waterproof fabric coating image, the degree of closure of the edge in the waterproof fabric coating image is determined. Based on the degree of closure of all edges in the waterproof fabric coating image and the LBP distribution entropy, the pinhole feature value of the waterproof fabric coating image is determined.
[0009] Based on the gray values of pixels in the waterproof fabric coating image and the position of the edges in the waterproof fabric coating image, identify the bubble center point and the bubble defect area of the bubble center point. Based on the difference in gray values of pixels in the bubble defect area of each bubble center point in the waterproof fabric coating image that are centrally symmetrical with respect to the bubble center point, determine the bubble feature value of the waterproof fabric coating image.
[0010] The quality assessment results of the waterproof fabric are obtained based on the crack feature values, pinhole feature values, and blister feature values of the waterproof fabric coating image.
[0011] Furthermore, the method for determining the local directional change rate of the target pixel is as follows:
[0012] The image of the waterproof fabric coating is converted to grayscale and edge detected to identify edges and edge pixels;
[0013] Record any pixel in the waterproof fabric coating image as the target pixel. Establish a local analysis window with a side length of the first preset length centered on the target pixel. Record the standard deviation of the gradient values of all edge pixels within the local analysis window as the local direction change rate of the target pixel.
[0014] Furthermore, the method for obtaining the local grayscale difference is as follows:
[0015] The range of gray values of all pixels within the local analysis window is recorded as the local gray-level difference of the target pixel.
[0016] Furthermore, the method for calculating the crack feature values of the waterproof fabric coating image is as follows:
[0017] The average local grayscale difference of all pixels in the waterproof fabric coating image is denoted as the grayscale difference of the waterproof fabric coating image.
[0018] The mean of the local orientation change rate of all pixels in the waterproof fabric coating image is denoted as the orientation change rate of the waterproof fabric coating image.
[0019] The positive correlation between the grayscale difference and the rate of change of direction of the waterproof fabric coating image is recorded as the crack feature value of the waterproof fabric coating image.
[0020] Furthermore, the method for determining the degree of edge closure in the waterproof fabric coating image is as follows:
[0021] The number of pixels contained in the region defined by the edge in the waterproof fabric coating image is denoted as the area of the region defined by the edge. The number of pixels contained within the smallest circumcircle of the region defined by the edge is denoted as the area of the smallest circumcircle of the region defined by the edge. The difference between the area of the smallest circumcircle of the region defined by the edge and the area of the region defined by the edge is denoted as the first difference of the edge. The ratio of the first difference of the edge to the area of the region defined by the edge is denoted as the first ratio of the edge. The negative correlation result of the first ratio of the edge is denoted as the degree of edge closure.
[0022] Furthermore, the method for calculating the pinhole feature values of the waterproof fabric coating image is as follows:
[0023] The normalized value of the mean LBP distribution entropy of all edges in the waterproof fabric coating image is denoted as the edge orientation uniformity of the waterproof fabric coating image.
[0024] The average degree of closure of all edges in the waterproof fabric coating image is denoted as the overall edge closure degree of the waterproof fabric coating image.
[0025] The mean value of the combined edge closure degree and edge direction uniformity of the waterproof fabric coating image is denoted as the pinhole feature value of the waterproof fabric coating image.
[0026] Furthermore, the method for identifying the center point of the bubbling is as follows:
[0027] All local maxima in the waterproof fabric coating image are recorded as candidate bubble center points;
[0028] Radial grayscale verification is performed on the candidate bubble center point: a verification window with a radius of a second preset length is established with the candidate bubble center point as the center, and the grayscale value in the verification window is verified to show a decreasing trend from the center to the surrounding area; candidate bubble center points that do not meet the trend are removed, and the remaining candidate bubble center points are recorded as bubble center points.
[0029] The method for identifying the blemish area at the center point of the blemish is as follows:
[0030] With the center point of the bubble as the center, construct an annulus with inner radius r1 and outer radius r2. The area corresponding to this annulus is denoted as the bubble defect area at the center point of the bubble, where 0 <r1<r2<=R;
[0031] in, , ,in and The preset proportional coefficient satisfies .
[0032] The method for determining the bubbling feature value of the waterproof fabric coating image is as follows:
[0033] Let any pixel on the outer radius of the bubble defect area at the bubble center point be the target outer radius point. Let the part of the line segment with the bubble center point and the target outer radius point as the two endpoints in the bubble defect area at the bubble center point be the radius of the ring between the bubble center point and the target outer radius point. Let the average gray value of all pixels on the ring radius between the bubble center point and the target outer radius point be the first gray value of the bubble center point and the target outer radius point.
[0034] Select the portion of the straight line passing through both the bubble center point and the target outer radius point within the bubble defect area at the bubble center point. Remove the radius of the annulus between the bubble center point and the target outer radius point, and record the remaining portion as the radius of the annulus opposite the bubble center point and the target outer radius point. Record the average gray value of all pixels on the annulus opposite the bubble center point and the target outer radius point as the second gray value of the bubble center point and the target outer radius point. Record the absolute value of the difference between the first gray value and the second gray value of the bubble center point and the target outer radius point as the brightness symmetry value of the bubble center point and the target outer radius point.
[0035] The mean value of the brightness symmetry feature value of all bubble center points in the waterproof fabric coating image is denoted as the bubble feature value of the waterproof fabric coating image.
[0036] Furthermore, the specific method for obtaining the quality assessment result of the waterproof fabric based on the crack feature values, pinhole feature values, and blister feature values of the waterproof fabric coating image includes:
[0037] The crack feature values, pinhole feature values, and blister feature values of the waterproof fabric coating image are arranged sequentially into a column vector to obtain the defect feature vector of the waterproof fabric coating image.
[0038] Input the defect feature vector of the waterproof fabric coating image of the waterproof fabric coating surface to be evaluated into the decision tree model to obtain the identification results of crack defects, pinhole defects and blister defects;
[0039] If the identification results for crack defects, pinhole defects, and bubbling defects are all "no defects", the waterproof fabric is determined to be without defects; otherwise, the waterproof fabric is determined to be defective.
[0040] The beneficial effects of this application are:
[0041] This application considers that cracks, pinholes, and bubbles in the waterproof fabric coating are all minor defects. First, based on the local contrast around each pixel, the features of crack defects are extracted to obtain crack feature values for the waterproof fabric coating image. Second, considering that pinhole defects are regularly shaped holes, the features of pinhole defects are extracted based on the closure degree and directional uniformity of the edges in the waterproof fabric coating image to obtain pinhole feature values. Then, considering that bubble defects are spherical reflections with a bright center and dark edges, the features of bubble defects are extracted based on the radial brightness variation and symmetry in the waterproof fabric coating image to obtain bubble feature values. Finally, based on the crack feature values, pinhole feature values, and bubble feature values of the waterproof fabric coating image, the quality assessment result of the waterproof fabric is obtained. This solves the problem that minor defects are easily submerged in the background texture, affecting the accuracy of the waterproof fabric quality assessment result, and improves the accuracy of identifying these three types of minor defects. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0043] Figure 1 This is a schematic diagram of a method for intelligent quality assessment of waterproof fabric provided in one embodiment of this application;
[0044] Figure 2 This is a flowchart illustrating the crack feature value acquisition process provided in one embodiment of this application. Detailed Implementation
[0045] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0046] Please see Figure 1 The diagram illustrates a flowchart of a smart quality assessment method for waterproof fabric provided in one embodiment of this application. The method includes the following steps:
[0047] Step S001: Acquire images of the waterproof fabric coating on the surface to be evaluated.
[0048] An industrial camera was used to capture images of the waterproof fabric coating surface to be evaluated. The images were then denoised and enhanced to reduce the impact of noise on the subsequent identification of cracks, pinholes, and bubbles.
[0049] This embodiment uses a median filtering algorithm to denoise the waterproof fabric coating image and a CLAHE histogram equalization algorithm to enhance the image. Both median filtering for denoising and histogram equalization for image enhancement are well-known techniques and will not be elaborated further. As other implementations, to achieve image denoising, implementers may use other methods from existing technologies such as Gaussian filtering and bilateral filtering; this application does not impose any special limitations.
[0050] At this point, the image of the waterproof fabric coating has been obtained.
[0051] Step S002: Denote any pixel in the waterproof fabric coating image as the target pixel, establish a local analysis window for the target pixel, and determine the local direction change rate and local grayscale difference of the target pixel based on the gradient difference of all edge pixels and the pixel value difference of all pixels in the local analysis window of the target pixel. Calculate the crack feature value of the waterproof fabric coating image based on the local direction change rate and local grayscale difference of all pixels in the waterproof fabric coating image.
[0052] Cracks, pinholes, and bubbles in the waterproof fabric coating are all minor defects, but they exhibit different characteristics. Cracks refer to fractures on the surface of the waterproof fabric coating, appearing as irregularly oriented, small connected regions in the image. Pinholes appear as regularly shaped areas with concentrated gradient distribution of edge pixels at their corresponding locations. Bubbles refer to protrusions on the waterproof fabric coating, appearing as spherical reflections with a bright center and dark edges. Therefore, features of cracks, pinholes, and bubbles are extracted separately, and feature fusion is performed based on these extracted features. The quality assessment result of the waterproof fabric is then obtained from the feature fusion result.
[0053] First, features of cracks and defects are extracted based on the local contrast around the pixel.
[0054] The waterproof fabric coating image is converted to grayscale to obtain a grayscale image of the waterproof fabric coating. Edge detection is then performed on the grayscale image of the waterproof fabric coating to obtain an edge image of the waterproof fabric coating. The edges and edge pixels in the waterproof fabric coating image can be identified through this edge image. In this embodiment, the Canny edge detection algorithm is used for edge detection. The use of the Canny edge detection algorithm for edge detection is a well-known technique and will not be described in detail here.
[0055] Let any pixel in the waterproof fabric coating image be designated as the target pixel, and establish a side length of [missing information] centered on the target pixel. The local window of the target pixel is obtained, the gradient value of all edge pixels within the local analysis window of the target pixel is calculated, and the standard deviation of the gradient values of all edge pixels within the local analysis window of the target pixel is denoted as the local direction change rate of the target pixel.
[0056] in, This represents the first preset length, which is 15 in this embodiment.
[0057] The same method can be used to obtain the local directional change rate of each pixel in the waterproof fabric coating image.
[0058] It should be noted that when edge pixels and edges are not identified in the waterproof fabric coating image, the local direction change rate of the target pixel is directly assigned to 0.
[0059] The local grayscale difference of the target pixel is determined based on the pixel value differences of all pixels within the local analysis window of the target pixel.
[0060] The range of gray values of all pixels within the local analysis window of the target pixel is denoted as the local gray-level difference of the target pixel.
[0061] The same method can be used to obtain the local grayscale difference of each pixel in the waterproof fabric coating image.
[0062] Crack feature values of the waterproof fabric coating image are calculated based on the local directional change rate and local grayscale difference of all pixels in the image.
[0063] The mean of the local grayscale differences of all pixels in the waterproof fabric coating image is denoted as the grayscale difference of the waterproof fabric coating image; the mean of the local directional change rate of all pixels in the waterproof fabric coating image is denoted as the directional change rate of the waterproof fabric coating image; the positive correlation between the grayscale difference and the directional change rate of the waterproof fabric coating image is denoted as the crack feature value of the waterproof fabric coating image.
[0064] The flowchart for obtaining crack feature values is as follows: Figure 2 As shown.
[0065] It is understood that a positive correlation is applied to the grayscale difference and the rate of change of direction, ensuring that the grayscale difference and the rate of change of direction are positively correlated with the crack characteristic value. It is understood that the positive correlation in this application refers to the relationship between the independent and dependent variables, where the independent variables are the grayscale difference and the rate of change of direction, and the dependent variable is the crack characteristic value. The positive correlation means that the dependent variable increases (decreases) as the independent variable increases (decreases), and can be an additive or multiplicative relationship.
[0066] Preferably, as an embodiment of this application, the normalized value of the product of the grayscale difference and the rate of change of direction of the waterproof fabric coating image is recorded as the crack feature value of the waterproof fabric coating image.
[0067] In this embodiment, the sigmoid function is used to calculate the normalized value. The sigmoid function is a well-known technique and will not be described in detail here. As other implementation methods, implementers can use other methods of the prior art, such as the tanh arctangent function.
[0068] In practical applications, as another implementation method, the average of the normalized values of the grayscale difference and the normalized values of the direction change rate of the waterproof fabric coating image is recorded as the crack feature value of the waterproof fabric coating image.
[0069] The larger the crack feature value of the waterproof fabric coating image, the greater the possibility of crack defects appearing in the waterproof fabric coating image.
[0070] At this point, the crack feature values of the waterproof fabric coating image are obtained.
[0071] Step S003: Based on the difference in LBP values of all edge pixels along the same edge in the waterproof fabric coating image, determine the LBP distribution entropy of the same edge; based on the difference in the number of pixels contained in the region divided by the edge and the circumcircle of the divided region in the waterproof fabric coating image, determine the degree of closure of the edge in the waterproof fabric coating image; based on the degree of closure of all edges and the LBP distribution entropy in the waterproof fabric coating image, determine the pinhole feature value of the waterproof fabric coating image.
[0072] Secondly, considering that the pinhole defects are regular-shaped holes, the features of the pinhole defects are extracted based on the closure degree and direction uniformity of the edges in the waterproof fabric coating image.
[0073] The edge direction uniformity of the waterproof fabric coating image is determined by the difference in LBP values of all edge pixels along the same edge in the waterproof fabric coating image.
[0074] Calculate the LBP value of each edge pixel in the waterproof fabric coating image. Based on the LBP values of all edge pixels along the same edge, generate an LBP histogram for that edge. Calculate the LBP distribution entropy of that edge based on the LBP histogram. The normalized value of the mean LBP distribution entropy of all edges in the waterproof fabric coating image is denoted as the edge direction uniformity of the waterproof fabric coating image.
[0075] Calculating the LBP value of a pixel, generating an LBP histogram, and calculating the LBP distribution entropy are all well-known techniques and will not be elaborated further.
[0076] The degree of edge closure in the waterproof fabric coating image is determined by the difference in the number of pixels contained in the region defined by the edge in the image and the circumcircle of the defined region.
[0077] The number of pixels contained in the region defined by the edge in the waterproof fabric coating image is denoted as the area of the region defined by the edge. The number of pixels contained within the smallest circumcircle of the region defined by the edge is denoted as the area of the smallest circumcircle of the region defined by the edge. The difference between the area of the smallest circumcircle of the region defined by the edge and the area of the region defined by the edge is denoted as the first difference of the edge. The ratio of the first difference of the edge to the area of the region defined by the edge is denoted as the first ratio of the edge. The negative correlation result of the first ratio of the edge is denoted as the degree of edge closure.
[0078] It is understood that the first ratio of the edges is negatively correlated, meaning that the first ratio of the edges is negatively correlated with the degree of closure of the edges. It is understood that the negative correlation in this application refers to the relationship between the independent variable and the dependent variable, where the independent variable is the first ratio of the edges and the dependent variable is the degree of closure of the edges. The negative correlation means that the dependent variable decreases (increases) as the independent variable increases (decreases), and can be an inverse relationship, a subtraction relationship, etc.
[0079] Preferably, as an embodiment of this application, the negative of the first ratio of the edges is taken as the exponent of an exponential function with the natural constant as the base, and the calculated value of the exponential function is recorded as the degree of closure of the edges.
[0080] Some other embodiments of this application may use the difference between the normalized value of the number 1 and the first ratio of the edge as the degree of edge closure.
[0081] It should be noted that when the edges of the waterproof fabric coating image cannot be divided into closed regions, the number of pixels contained in the regions divided by the edges in the waterproof fabric coating image cannot be calculated, and the degree of closure of the edges is directly assigned to 0.
[0082] The pinhole feature values of the waterproof fabric coating image are determined based on the degree of closure of all edges in the image and the uniformity of edge direction.
[0083] The average degree of closure of all edges in the waterproof fabric coating image is denoted as the overall edge closure degree of the waterproof fabric coating image. The average value of the overall edge closure degree and the edge direction uniformity of the waterproof fabric coating image is denoted as the pinhole feature value of the waterproof fabric coating image.
[0084] The larger the pore feature value of the waterproof fabric coating image, the greater the possibility of pore defects appearing in the waterproof fabric coating image.
[0085] It should be noted that when edge pixels and edges are not identified in the waterproof fabric coating image, the feature value of the small hole in the waterproof fabric coating image is directly assigned to 0.
[0086] At this point, the pore feature values of the waterproof fabric coating image are obtained.
[0087] Step S004: Based on the grayscale values of pixels in the waterproof fabric coating image and the position of the edges in the waterproof fabric coating image, identify the bubble center point and the bubble defect area of the bubble center point. Based on the difference in grayscale values of pixels within the bubble defect area of each bubble center point in the waterproof fabric coating image that are centrally symmetrical to the bubble center point, determine the bubble feature value of the waterproof fabric coating image.
[0088] Then, considering that the location of the blistering defect is a spherical reflection with a bright center and dark edges, the features of the blistering defect are extracted based on the radial brightness variation and symmetry in the waterproof fabric coating image.
[0089] The LoG operator is used to process the waterproof fabric coating image to identify local maxima points in the image, and these local maxima points are recorded as candidate bubble center points. Radial grayscale verification is then performed on the candidate bubble center points: a verification window with a radius of a second preset length is established with the candidate point as the center, and the grayscale values within the verification window are verified to show a decreasing trend from the center to the surrounding areas. Candidate bubble center points that do not meet the trend are eliminated, and the remaining candidate bubble center points are recorded as bubble center points.
[0090] Specifically, the verification process may include: taking the candidate bubble center point as the center, within the verification window, extracting the gray value sequences from the center point outward along multiple preset radial directions (for example, along 8 directions); then, performing trend analysis on the gray value sequences in each direction to determine whether their gray values steadily decrease as the distance from the center point increases (for example, by analyzing their correlation coefficients or first-order differences); finally, counting the number of radial directions that satisfy the downward trend. When this number is greater than a preset proportion threshold (for example, more than 75% of the directions satisfy), it is determined that the candidate point is a real bubble center point and is retained, otherwise it is excluded.
[0091] Taking the bubble center point as the center of a circle, radially search for local minimum points of gray values (i.e., the dark edges of the bubble) around; record the average Euclidean distance between the bubble center point and the local minimum points as the bubble radius R of the bubble center point.
[0092] Taking this bubble center point as the center of a circle, establish a circular ring with an inner radius r1 and an outer radius r2 (where 0 < r1 < r2 <= R), and record the area corresponding to this circular ring as the bubble defect area of the bubble center point.
[0093] Among them, the inner radius r1 and the outer radius r2 are adaptively determined according to the bubble radius R, aiming to ensure that the analysis area (circular ring) can effectively capture the spherical-like reflection characteristics of the bubble while avoiding interference areas. Specifically, the inner radius r1 aims to avoid the specular reflection highlight area that may exist near the bubble center point, and the outer radius r2 aims to ensure that the analysis area is located inside the bubble radius R (i.e., the dark edge of the gray local minimum).
[0094] Preferably, the inner radius r1 and the outer radius r2 can be set as proportional values of the bubble radius R, that is , , where and are preset proportionality coefficients, satisfying . In this embodiment, it can be set that has a value range of [0.2, 0.4], has a value range of [0.7, 0.9].
[0095] Among them, the LoG operator is the Laplacian of Gaussian operator. Using the LoG operator to identify local maximum points in an image is a well-known technique and will not be elaborated here.
[0096] It can be understood that the bubble center point corresponds to the brightest point at the center of the bubble defect position, and the bubble defect area (circular ring) of the bubble center point corresponds to the main part of the spherical-like reflection of the bubble defect position.
[0097] Let any pixel on the outer radius of the bubble defect area at the bubble center point be the target outer radius point. Let the portion of the line segment with the bubble center point and the target outer radius point as its two endpoints within the bubble defect area at the bubble center point be the radius of the annulus between the bubble center point and the target outer radius point. Let the average gray value of all pixels on the annulus radius between the bubble center point and the target outer radius point be the first gray value of the bubble center point and the target outer radius point. Select the portion of the straight line passing through both the bubble center point and the target outer radius point within the bubble defect area at the bubble center point, remove the portion containing the annulus radius between the bubble center point and the target outer radius point, and let the remaining portion be the radius of the annulus opposite the bubble center point and the target outer radius point. Let the average gray value of all pixels on the annulus radius opposite the bubble center point and the target outer radius point be the second gray value of the bubble center point and the target outer radius point. The absolute value of the difference between the first grayscale mean and the second grayscale mean between the bubble center point and the target outer radius point is recorded as the brightness symmetry value between the bubble center point and the target outer radius point.
[0098] Using the same method, we can obtain the brightness symmetry value of any pixel on the outer radius of the bubble defect area at the bubble center point. In other words, every pixel on the outer radius of the bubble defect area at the bubble center point has a corresponding brightness symmetry value. The normalized value of the mean of the brightness symmetry values of all pixels on the outer radius of the bubble defect area at the bubble center point is denoted as the first mean of the bubble center point. The difference between the number 1 and the first mean of the bubble center point is denoted as the brightness symmetry feature value of the bubble center point.
[0099] The mean value of the brightness symmetry feature value of all bubble center points in the waterproof fabric coating image is denoted as the bubble feature value of the waterproof fabric coating image.
[0100] The larger the bubbling feature value of the waterproof fabric coating image, the greater the possibility of bubbling defects appearing in the waterproof fabric coating image.
[0101] It should be noted that if the center point of the bubble is not identified in the image of the waterproof fabric coating, the bubble feature value of the waterproof fabric coating image should be directly assigned to 0.
[0102] At this point, the bubbling feature values of the waterproof fabric coating image are obtained.
[0103] Step S005: Obtain the quality assessment results of the waterproof fabric based on the crack feature values, pinhole feature values, and blister feature values of the waterproof fabric coating image.
[0104] The crack feature values, pinhole feature values, and blister feature values of the waterproof fabric coating image are arranged sequentially into a column vector to obtain the defect feature vector of the waterproof fabric coating image.
[0105] The waterproof fabric coating surface was manually inspected for cracks, pinholes, and bubbles. One thousand waterproof fabric coating surfaces were obtained, each containing one, two, or all three defects, and one surface without any defects. Using the same method, images of the waterproof fabric coating and defect feature vectors from these 1000 surfaces were obtained. These feature vectors were then used to create a dataset, which was divided into training, validation, and test sets in a 6:2:2 ratio. A decision tree model was constructed based on these sets. The defect feature vectors were input into the decision tree model to obtain the identification results for cracks, pinholes, and bubbles. Specifically, the identification results for cracks, pinholes, and bubbles were either "defect present" or "no defect present."
[0106] The defect feature vector of the waterproof fabric coating image to be evaluated is input into the decision tree model to obtain the identification results of crack defects, pinhole defects, and blister defects. When the identification results of crack defects, pinhole defects, and blister defects are all "no defects", the waterproof fabric is determined to be defect-free; when the identification results of crack defects, pinhole defects, and blister defects are all "defects present", the waterproof fabric is determined to be defective.
[0107] At this point, the quality assessment results for the waterproof fabric have been obtained.
[0108] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.
Claims
1. A waterproof cloth quality intelligent evaluation method, characterized in that, The method includes the following steps: Acquire images of the waterproof fabric coating on the surface to be evaluated; Denote any pixel in the waterproof fabric coating image as the target pixel, establish a local analysis window for the target pixel, and determine the local direction change rate and local gray level difference of the target pixel based on the pixel value difference of all pixels in the local analysis window and the gradient difference of all edge pixels. Calculate the crack feature value of the waterproof fabric coating image based on the local gray level difference and local gray level difference of all pixels in the waterproof fabric coating image. Based on the difference in LBP values of all edge pixels along the same edge in the waterproof fabric coating image, the LBP distribution entropy of the same edge is determined. Based on the difference in the number of pixels contained in the region divided by the edge and the outer circle of the divided region in the waterproof fabric coating image, the degree of closure of the edge in the waterproof fabric coating image is determined. Based on the degree of closure of all edges in the waterproof fabric coating image and the LBP distribution entropy, the pinhole feature value of the waterproof fabric coating image is determined. Based on the gray values of pixels in the waterproof fabric coating image and the position of the edges in the waterproof fabric coating image, identify the bubble center point and the bubble defect area of the bubble center point. Based on the difference in gray values of pixels in the bubble defect area of each bubble center point in the waterproof fabric coating image that are centrally symmetrical with respect to the bubble center point, determine the bubble feature value of the waterproof fabric coating image. Based on the crack feature values, pinhole feature values, and blister feature values of the waterproof fabric coating image, the quality assessment results of the waterproof fabric are obtained. The method for determining the local directional change rate of the target pixel is as follows: The image of the waterproof fabric coating is converted to grayscale and edge detected to identify edges and edge pixels; Record any pixel in the waterproof cloth coating image as the target pixel. Establish a local analysis window with a side length of the first preset length centered on the target pixel. Record the standard deviation of the gradient values of all edge pixels in the local analysis window as the local direction change rate of the target pixel. The method for obtaining the local grayscale difference is as follows: The range of gray values of all pixels within the local analysis window is recorded as the local gray-level difference of the target pixel.
2. The method of claim 1, wherein, The method for calculating the crack feature value of the waterproof fabric coating image is as follows: The average local grayscale difference of all pixels in the waterproof fabric coating image is denoted as the grayscale difference of the waterproof fabric coating image. The mean of the local orientation change rate of all pixels in the waterproof fabric coating image is denoted as the orientation change rate of the waterproof fabric coating image. The normalized value of the product of the grayscale difference and the rate of change of direction of the waterproof fabric coating image is denoted as the crack feature value of the waterproof fabric coating image.
3. The method of claim 1, wherein, The method for determining the degree of edge closure in the waterproof fabric coating image is as follows: The number of pixels contained in the region defined by the edge in the waterproof fabric coating image is denoted as the area of the region defined by the edge. The number of pixels contained in the smallest circumcircle of the region defined by the edge is denoted as the area of the smallest circumcircle of the region defined by the edge. The difference between the area of the smallest circumcircle of the region defined by the edge and the area of the region defined by the edge is denoted as the first difference of the edge. The ratio of the first difference of the edge to the area of the region defined by the edge is denoted as the first ratio of the edge. The difference between the number 1 and the normalized value of the first ratio of the edge is denoted as the degree of closure of the edge.
4. The method of claim 1, wherein, The method for calculating the pinhole feature value of the waterproof fabric coating image is as follows: The normalized value of the mean LBP distribution entropy of all edges in the waterproof fabric coating image is denoted as the edge orientation uniformity of the waterproof fabric coating image. The average degree of closure of all edges in the waterproof fabric coating image is denoted as the overall edge closure degree of the waterproof fabric coating image. The mean value of the combined edge closure degree and edge direction uniformity of the waterproof fabric coating image is denoted as the pinhole feature value of the waterproof fabric coating image.
5. The intelligent quality assessment method for waterproof fabric according to claim 1, characterized in that, The method for identifying the center point of the bubbling is as follows: All local maxima in the waterproof fabric coating image are recorded as bubble center points.
6. The method of claim 1, wherein, The method for identifying the blemish area at the center point of the blemish is as follows: When the center point of a bubble is not within the closed area defined by the edge of the waterproof fabric coating image, the center point of the bubble is removed. Calculate the Euclidean distance between the bubble center point and each edge pixel in the closed region corresponding to the bubble center point. Record the minimum Euclidean distance as the first radius of the bubble center point and the maximum Euclidean distance as the second radius of the bubble center point. With the center point of the bubble as the center, the inner radius as the first radius of the bubble center point, and the outer radius as the second radius of the bubble center point, establish a ring. The area corresponding to the ring is recorded as the bubble defect area of the bubble center point.
7. The method of claim 1, wherein, The method for determining the bubbling feature value of the waterproof fabric coating image is as follows: Let any pixel on the outer radius of the bubble defect area at the bubble center point be the target outer radius point. Let the part of the line segment with the bubble center point and the target outer radius point as the two endpoints in the bubble defect area at the bubble center point be the radius of the ring between the bubble center point and the target outer radius point. Let the average gray value of all pixels on the ring radius between the bubble center point and the target outer radius point be the first gray value of the bubble center point and the target outer radius point. Select the portion of the straight line passing through both the bubble center point and the target outer radius point within the bubble defect area at the bubble center point, and remove the radius of the ring between the bubble center point and the target outer radius point. Record the remaining portion as the radius of the ring opposite the bubble center point and the target outer radius point. Record the average gray value of all pixels on the ring radius opposite the bubble center point and the target outer radius point as the second gray value of the bubble center point and the target outer radius point. The absolute value of the difference between the first gray-scale mean and the second gray-scale mean between the bubble center point and the target outer radius point is recorded as the brightness symmetry value between the bubble center point and the target outer radius point. The normalized value of the mean of the brightness symmetry value of all pixels on the outer radius of the bubble defect area at the bubble center point and the bubble center point is recorded as the first mean of the bubble center point. The difference between the number 1 and the first mean of the bubble center point is recorded as the brightness symmetry feature value of the bubble center point. The maximum value of the brightness symmetry feature value of all bubble center points in the waterproof fabric coating image is recorded as the bubble feature value of the waterproof fabric coating image.
8. The method of claim 1, wherein, The method for obtaining the quality assessment result of the waterproof fabric based on the crack feature values, pinhole feature values, and blister feature values of the waterproof fabric coating image includes the following specific methods: The crack feature values, pinhole feature values, and blister feature values of the waterproof fabric coating image are arranged sequentially into a column vector to obtain the defect feature vector of the waterproof fabric coating image. Input the defect feature vector of the waterproof fabric coating image of the waterproof fabric coating surface to be evaluated into the decision tree model to obtain the identification results of crack defects, pinhole defects and blister defects; If the identification results for crack defects, pinhole defects, and bubbling defects are all "no defects", the waterproof fabric is determined to be without defects; otherwise, the waterproof fabric is determined to be defective.
Citation Information
Patent Citations
Waterproof cloth quality data processing and identification method
CN114998346A
Cloth defect detection and classification method for garment production
CN120410984A