Quality inspection method for plastic lace

Abstract

The invention provides a plastic lace quality inspection method, which comprises the following steps: acquiring a plastic lace thermal deformation defect image, and carrying out image preprocessing on the image, the preprocessing comprising denoising and enhancement, so as to obtain defect image data; extracting related data of a bending defect, a distortion defect and a shrinkage defect in the defect image data to obtain a defect area, performing shape feature extraction on the related data, and constructing the extracted shape features into a feature vector to obtain a shape feature vector of the thermal deformation defect; fusing the shape feature vector and the size feature vector of the thermal deformation defect to construct a comprehensive feature vector of the thermal deformation defect; the method comprises the following steps: inputting a comprehensive feature vector of a thermal deformation defect into a preset thermal deformation defect identification model for training, learning a preset labeled sample, and establishing a discrimination function between a thermal deformation defect type and the comprehensive feature vector;

Description

Technical Field

[0001] The invention relates to the field of information technology, and in particular to a quality inspection method for plastic lace. Background Art

[0002] In the detection of thermal deformation defects of plastic lace, there are difficulties in extracting shape features and size features. First, shape feature extraction requires accurate quantification of the radius and angle of the bending defect, the torsion direction and degree of the twisting defect, the proportion and direction of the shrinkage defect, and the degree of the expansion defect. However, due to the complex surface texture and irregular edge contour of the plastic lace, the feature extraction accuracy is insufficient. Secondly, the measurement of size features requires accurate acquisition of the length, width and area information of the defect area, but it is affected by image noise, uneven lighting and shadow interference, resulting in unstable measurement results. These two types of features are interrelated. The accurate extraction of shape features depends on the accurate measurement of size features, and the reliable acquisition of size features is subject to the correct identification of shape features. Therefore, it is necessary to perform denoising and enhancement in the image preprocessing stage to improve the accuracy of edge detection and contour extraction. At the same time, a fusion model of shape features and size features is established to achieve mutual correction and optimization of feature extraction, thereby improving the overall performance of thermal deformation defect detection. Summary of the invention

[0003] The present invention provides a quality inspection method for plastic lace, which mainly comprises: The thermal deformation defect image of plastic lace is obtained, and image preprocessing is performed on the image, including denoising and enhancement, to obtain defect image data; relevant data of bending defects, twisting defects and shrinkage defects in the defect image data are extracted to obtain the defect area, and shape features are extracted from the relevant data, and the extracted shape features are constructed into feature vectors to obtain the shape feature vectors of the thermal deformation defects; the geometric size parameters of the defect area are measured, the total area of ​​the defect area and the area ratio relative to the entire image are calculated as relative area features, and the geometric size parameters of the defect area are combined into feature vectors to obtain the size feature vectors of the thermal deformation defects; the shape feature vectors and size feature vectors of the thermal deformation defects are fused to construct the comprehensive feature vectors of the thermal deformation defects; the comprehensive feature vectors of the thermal deformation defects are combined into a feature vector to obtain the size feature vectors of the thermal deformation defects; The eigenvector is input into a preset thermal deformation defect recognition model for training, and the preset labeled samples are learned to establish a discriminant function between the thermal deformation defect type and the comprehensive eigenvector; the discriminant function between the established thermal deformation defect type and the comprehensive eigenvector is used to extract the comprehensive eigenvector from the input plastic lace thermal deformation defect image, and the thermal deformation defect recognition result is obtained according to the output result. The thermal deformation defect recognition result is used to determine the defect type through the discriminant function, and the defect types include bending, twisting and shrinkage; the bending defect arc radius threshold, twisting defect torsion angle threshold and shrinkage defect shrinkage ratio threshold corresponding to the defect type are obtained. If the identified defect arc radius is less than the threshold and the torsion angle is greater than the threshold and the shrinkage ratio is greater than the threshold, the plastic lace is judged to be unqualified, otherwise the quality is judged to be qualified.

[0004] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: The present invention discloses a quality inspection method for plastic lace. The method aims at the common thermal deformation defect problem in the production of plastic lace, and realizes automatic identification and quality judgment through image processing and machine learning technology. Specifically, the present invention first pre-processes and extracts features of the defect image, and constructs a comprehensive feature vector containing shape and size information. Then, the defect type and parameter measurement of the input image are performed using a pre-trained recognition model. Finally, according to the set threshold standard, it is automatically determined whether the product quality is qualified. The technical solution of the present invention can effectively improve the quality control efficiency and accuracy of plastic lace production, reduce the workload of manual inspection, and is of great significance to improving the product yield rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Figure 1 The present invention is a flow chart of a quality inspection method for plastic lace.

[0006] Figure 2 It is a schematic diagram of a quality inspection method for plastic lace according to the present invention.

[0007] Figure 3It is another schematic diagram of a quality inspection method for plastic lace according to the present invention. DETAILED DESCRIPTION

[0008] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0009] like Figure 1-3 In this embodiment, a quality inspection method for plastic lace may specifically include: Step S101, obtaining a plastic lace thermal deformation defect image, performing image preprocessing on the image, the preprocessing including denoising and enhancement, to obtain defect image data.

[0010] For the original image of thermal deformation defects of plastic lace, the noise points in the original image are filtered out by median filtering, and the filtered image is processed by bilateral filter to obtain a second image; a block adaptive histogram equalization operation is performed on the second image, and a nonlinear mapping transformation is performed based on the brightness characteristics of the local area of ​​the image to obtain a third image; for the third image, the Canny edge detection operator is used to extract the edge contour of the thermal deformation defect, the edge contour is processed by morphological opening operation, and the processed edge is segmented by grayscale threshold to obtain a fourth image; for the defect contour area segmented in the fourth image, the defect contour area is marked, and if the area of ​​the connected domain of the defect contour area is greater than the area threshold and the grayscale mean reaches the preset grayscale mean, the defect contour area is determined to be a target defect.

[0011] Specifically, the image of thermal deformation defect of plastic lace is collected according to the image acquisition device, and the salt and pepper noise points in the original image are filtered out by the median filter of the 3x3 template size. The second image is obtained by using the bilateral filter of the 5x5 window size to maintain the image edge structure information and filter out the Gaussian white noise. The second image is subjected to the adaptive histogram equalization operation with the 8x8 block size, and the local cumulative histogram of the gray value in each block image is calculated. The image contrast is enhanced by the nonlinear mapping transformation based on the brightness characteristics of the local area of ​​the image to obtain the third image. The edge contour of the thermal deformation defect is extracted by the Canny edge detection operator for the third image, and the small edge interference points are removed by the morphological opening operation according to the 3x3 circular structure element. The background interference is removed by the gray threshold of 125 for the edge image, and the image is segmented to obtain the fourth image. For the defect contour area segmented in the fourth image, the connected domain is marked for the defect area by the 8-neighborhood marking method, and the connected domain area is screened according to the connected domain area greater than 100 pixels, and the defect area is binarized according to the defect gray mean of 250 to obtain the defect image data. The defects of plastic lace during thermal deformation are mainly manifested as physical changes such as fracture, wrinkles, and deformation. These defects show obvious differences in grayscale values ​​from normal areas in the image. When collecting the original image, due to uneven ambient lighting and sensor noise interference, the image will be mixed with salt and pepper noise and Gaussian noise. Salt and pepper noise is manifested as randomly distributed black and white pixels. The median value in the neighborhood of each pixel is replaced by the original pixel value through a 3x3 median filter window to effectively remove such mutation noise points. Gaussian noise is manifested as random fluctuations in pixel values. A 5x5 bilateral filter is used to smooth the noise while maintaining edge jumps. 8x8 block adaptive histogram equalization is used in image enhancement, and the cumulative histogram is calculated for each block. If the maximum grayscale value in a block is 200, the minimum grayscale value is 50, the total number of pixels in the block is 64, and the cumulative number of pixels with a grayscale value of 125 is 32, then the new value of the grayscale value after mapping is 125. This local adaptive enhancement method avoids the problem of over-enhancement caused by global histogram equalization. The Canny operator is used to extract the edge features of thermal deformation defects. The edge pixel position is determined by calculating the image gradient amplitude and direction. The edge of the defect area often presents an irregular shape. A 3x3 circular structure element is used for morphological opening operation to remove small noise points while maintaining the main edge structure. The image segmentation threshold is set to 125, and the background area pixels below the threshold are set to 0 to retain the defect edge contour information. The segmented defect area is marked with an 8-neighborhood connected domain to determine the connectivity of each pixel with its surrounding 8 pixels. If the area of ​​the connected domain of a defect area is less than 100 pixels, it is considered to be noise interference and is removed. The defect area often presents a higher gray value feature. The gray threshold is set to 250 for binarization, and the area above the threshold is determined as a defect area.In practical applications, the area of ​​thermal deformation defects in plastic lace is generally between 200 and 800 pixels, and the grayscale values ​​are concentrated between 230 and 255. These features can serve as an important basis for defect detection.

[0012] Step S102, extract relevant data of bending defects, twisting defects and shrinkage defects in the defect image data to obtain the defect area, and extract shape features of the relevant data, construct the extracted shape features into a feature vector, and obtain the shape feature vector of the thermal deformation defect.

[0013] The connected domain area of ​​the edge contour of the defect area is calculated, and the Hough circle transform is applied to the edge contour curve to obtain the arc radius and arc angle of the curved defect area, and the shape features are extracted from the coordinate sequence of the contour curve to obtain the second image; for the marked area in the second image, the image moment of inertia is used to calculate the main axis direction of the area; if the gray value of the marked area shows periodic changes in the main axis direction, it is determined to be a twisted defect area; the angle between the main axis direction of the twisted defect area and the horizontal direction is obtained to obtain the torsion angle; according to the fluctuation amplitude of the gray value of the twisted defect area in the main axis direction, the degree of torsion is calculated; the shape features of the contour of the twisted defect area are extracted to obtain to a third image; extract a marked area from the third image, calculate the coordinates of the center of mass of the area, and obtain the position of the shrinkage center point; determine the shrinkage direction by the change in the distance from the boundary point of the extracted marked area to the center of mass; calculate the shrinkage ratio by the ratio of the area of ​​the area after shrinkage to the area of ​​the initial area; extract the shape features of the shrinkage defect according to the change in the curvature of the contour points of the extracted marked area, and obtain a fourth image; for the shape features extracted from the fourth image, use the minimum and maximum value normalization method to process the bending defect features, the twisting defect features and the shrinkage defect features data respectively, and construct a thermal deformation defect shape feature vector according to the feature correlation weight to obtain the shape feature vector of the thermal deformation defect.

[0014] Specifically, after the defect area of ​​the image segmentation has been extracted by the Canny edge detection operator, the 8-neighborhood labeling method is used to calculate the area of ​​the connected domain for the edge contour, and the Hough circle transform is used to calculate the arc radius and arc angle of the curved defect area for the contour curve. The arc length value is obtained by the curve arc length calculation formula, and the shape features are extracted from the contour curve coordinate sequence to obtain the second image. , this formula is used to calculate the arc length of the curve y=f(x) from x=a to x=b, where dy / dx is the derivative of the curve, which represents the slope of the curve at any point. For the marked area in the second image, the image moment of inertia is used to calculate the direction of the main axis of the area. If the grayscale value of the area shows periodic changes in the main axis direction, it is determined to be a twisted defect area. The torsion angle is obtained by the angle between the main axis direction and the horizontal direction. The degree of torsion is calculated according to the fluctuation amplitude of the grayscale value of the area in the main axis direction. The shape features of the twisted defect contour are extracted to obtain the third image. The marked area is extracted from the third image, the coordinates of the center of mass of the area are calculated to obtain the position of the shrinkage center point, and the shrinkage direction is determined by the change in the distance from the boundary point of the area to the center of mass. The shrinkage ratio is calculated by the ratio of the area of ​​the area after shrinkage to the area of ​​the initial area. The shape features of the shrinkage defect are extracted according to the change in the curvature of the area contour point to obtain the fourth image. The shrinkage ratio is calculated by the following formula: , A final Represents the area after contraction, A initial Represents the area of ​​the initial region. For the shape features extracted from the fourth image, the bending defect feature, the twisting defect feature, and the shrinkage defect feature data are processed by minimum and maximum value normalization, and the weights of the normalized bending defect feature, the twisting defect feature, and the shrinkage defect feature are calculated respectively. The shape feature vector of the thermal deformation defect is constructed according to the feature correlation weight to obtain the shape feature vector of the thermal deformation defect.

[0015] , w bend represents the weight of the bending defect feature, M represents the number of samples, and y i represents the label of the i-th sample, x i , bend represents the bending defect characteristic value of the i-th sample.

[0016] , w twist represents the weight of the distortion defect feature, M represents the number of samples, and y i represents the label of the i-th sample, x i , twist represents the twist defect characteristic value of the i-th sample.

[0017] , w shrink represents the weight of the shrinkage defect feature, M represents the number of samples, and y i represents the label of the i-th sample, x i, shrink represents the shrinkage defect characteristic value of the i-th sample. Bending, twisting and shrinkage defects in plastic lace thermal deformation defects have unique shape characteristics. Bending defects are manifested as arc deformation of the lace locally. The contour extracted by Canny edge detection shows obvious arc characteristics. The arc radius and angle can be calculated by Hough circle transform. When the lace is bent 90 degrees, its arc radius is usually between 5 and 15 mm, and the arc length accounts for 15% to 25% of the entire contour length. These characteristic values ​​constitute the quantitative indicators of bending defects. Twist defects are reflected in the image as periodic changes in grayscale values, and the main axis direction is obtained by calculating the image moment of inertia. In actual cases, when the plastic lace is twisted 45 degrees, the grayscale value along the main axis direction shows 4 to 6 cycles of fluctuations, and the fluctuation amplitude is between 50 and 80 gray levels. The degree of twisting is determined by the frequency and amplitude of grayscale fluctuations. The higher the grayscale fluctuation frequency and the larger the amplitude, the more serious the twisting. Shrinkage defects are characterized by regional area changes and shape distortion. The centroid of the defective area is calculated as the shrinkage center point. The standard deviation of the distance from the boundary point to the centroid of normal plastic lace is in the range of 2 to 3 mm, while this value increases to 4 to 6 mm after shrinkage. The shrinkage ratio is calculated by the area ratio. The area ratio of slight shrinkage is between 0.85 and 0.95, and severe shrinkage drops below 0.7. The three defect features are normalized and fused to form a feature vector. Taking a plastic lace sample as an example, the bending defect feature includes an arc radius of 12 mm, an arc angle of 75 degrees, and an arc length of 18 mm; the twisting defect feature includes a torsion angle of 35 degrees, a fluctuation frequency of 5 cycles, and a grayscale fluctuation amplitude of 65; the shrinkage defect feature includes a shrinkage ratio of 0.82 and a standard deviation of 4.8 mm from the boundary point. These eigenvalues ​​are mapped to the range of 0 to 1 through minimum and maximum value normalization, and then the feature weights are calculated according to the linear discriminant function to construct a comprehensive feature vector reflecting the degree of thermal deformation.

[0018] , v represents the comprehensive feature vector, w i represents the weight of the i-th feature, x i represents the i-th eigenvector, and N represents the number of features. In the feature extraction process, the edge detection threshold, main axis direction calculation, and regional area statistics directly affect the accuracy of the eigenvalue. Practice shows that the low threshold of the Canny operator is set to 50 and the high threshold is set to 150, which can effectively extract the edge contour of the plastic lace. The image moment of inertia calculation uses a 9x9 neighborhood window to accurately obtain the main axis direction of the distorted area. When calculating the shrinkage center point, a 3x3 Gaussian weight is used to smooth the coordinates of the center point to improve the stability of the center point position.

[0019] Step S103, measuring the geometric size parameters of the defect area, calculating the total area of ​​the defect area and the area ratio relative to the entire image as relative area features, combining the geometric size parameters of the defect area into a feature vector, and obtaining the size feature vector of the thermal deformation defect.

[0020] The Sobel operator is used to obtain the edge contour point set of the defect area, the projection length and width are calculated according to the edge contour point set, the actual contour length is obtained by calculating the cumulative value of the contour point distance, and the actual width is obtained by the vertical distance between the contour point set and the projection axis; the number of pixels occupied by the defect area is extracted, the relative area ratio is calculated by the ratio of the number of pixels to the total number of pixels in the image, and the mean filter is used to remove abnormal values ​​of the measurement data; the minimum and maximum value normalization is used to process the length, width and area values ​​of the defect area, the feature weight is calculated by the Pearson correlation coefficient of the length, width and area values ​​of the defect area, and the size feature vector of the thermal deformation defect is obtained.

[0021] Specifically, the defect edge contour point set is obtained by using the Sobel edge detection operator according to the defect area obtained by image segmentation, the minimum circumscribed rectangle is used to calculate the projection length and width of the defect area in the horizontal and vertical directions for the edge contour point set, the area of ​​the defect area is calculated by counting all the pixels in the edge contour point set and multiplying by the actual area of ​​a single pixel, and the coordinates of the defect edge contour point set are recorded to obtain the fifth image. For the defect edge point set recorded in the fifth image, the four-neighborhood point tracking algorithm is used to calculate the cumulative value of the distance between adjacent contour points to obtain the actual contour length of the defect area, the actual length value is calculated by the pre-calibrated pixel and actual size conversion ratio, and the actual width of the defect area is obtained according to the vertical distance between the contour point set and the projection axis to obtain the sixth image. The area data of the defect area is extracted from the sixth image, the relative area ratio is calculated by the ratio of the number of pixels occupied by the defect area to the total number of pixels in the image, the actual area is calibrated according to the lens parameters and working distance of the image acquisition device, and the abnormal value of the measured data is removed based on the mean filter to obtain the seventh image. For the geometric size features extracted from the seventh image, the minimum and maximum values ​​are normalized to process the length, width, area, and relative area values ​​of the defect area. The feature weight is calculated according to the Pearson correlation coefficient of the feature data. The thermal deformation defect size feature vector is constructed by serial combination to obtain the thermal deformation defect feature data. The size measurement of the thermal deformation defect area of ​​plastic lace needs to be analyzed from multiple dimensions such as edge contour, geometric size, and area features. Sobel edge detection obtains edge contour points by calculating the grayscale difference of the image in the horizontal and vertical directions. For the thermal deformation defect image, the horizontal operator kernel size is 3x3, and the vertical operator kernel size is 3x3. When the grayscale jump of the image edge exceeds the threshold of 125, it is determined as an edge point. For typical lace thermal deformation defects, its minimum bounding rectangle reflects the overall size range of the defect area. In actual measurement, in a 1000x1000 pixel image, the normal plastic lace width is about 200 pixels. After thermal deformation, the minimum bounding rectangle width of the defect area increases to 250 to 300 pixels. By using the pre-calibrated pixel size ratio of 0.05 mm / pixel, the actual width of the defect area can be converted to between 12.5 and 15 mm. When the four-neighborhood point tracking is used to calculate the actual length of the defect contour, the point with the largest grayscale value is selected from the edge point set as the starting point, and the adjacent edge points are searched in the order of top, right, bottom, and left. For a typical arc-shaped thermal deformation defect, the total length of its contour is about 400 to 600 pixels, and the actual length is between 20 and 30 mm. The actual width of the defect area is determined by the maximum vertical distance from the contour point to the horizontal projection axis, which is generally between 8 and 12 mm. The area calculation of the defect area is based on pixel counts, and a moderate thermal deformation defect area occupies 40,000 to 60,000 pixels. In a 1000x1000 image, the relative area ratio is between 4% and 6%.When the focal length of the camera lens is set to 50 mm and the working distance is maintained at 300 mm, the actual area corresponding to a single pixel is 0.0025 square millimeters. The actual area of ​​the defect area is calculated to be between 100 and 150 square millimeters. The normalization of geometric dimension features adopts the minimum and maximum value method. Taking the defect length as an example, if the maximum length in the sample data is 30 mm and the minimum length is 15 mm, and the measured length of a defect is 20 mm, then its normalized value is 0.33. Feature weight calculation shows that in the identification of thermal deformation defects, the correlation coefficient between the relative area feature and the defect length reaches 0.85, indicating that the two have a strong correlation, while the correlation coefficient of the width feature is lower, only 0.45. In the final constructed feature vector, the relative area and length features obtain higher weights, 0.4 and 0.35 respectively, while the width feature weight is 0.25.

[0022] Step S104, the shape feature vector and the size feature vector of the thermal deformation defect are merged to construct a comprehensive feature vector of the thermal deformation defect.

[0023] The shape feature vector and the size feature vector in the defect image are obtained, and Z-score normalization processing is performed on the shape feature vector and the size feature vector, and abnormal data is eliminated by using the standard deviation principle to obtain a second feature vector; the Pearson correlation coefficient is calculated according to the feature value in the second feature vector, if the correlation coefficient is a positive value, the positive correlation feature is retained, if the correlation coefficient is a negative value, the negative correlation feature is retained, and weights are assigned to the feature values ​​by the absolute value of the correlation coefficient to obtain a third feature vector; weighted average calculation is performed on the feature values ​​to which weights are assigned in the third feature vector, and maximum and minimum value normalization is used to perform numerical mapping on the weighted summation result to obtain a fourth feature vector; principal component analysis is performed on the normalized feature values ​​in the fourth feature vector, the number of principal components is screened by the cumulative contribution rate threshold, and a comprehensive feature vector is constructed according to the principal component scores to obtain a comprehensive feature vector of the thermal deformation defect.

[0024] Specifically, according to the shape feature vector and size feature vector extracted from the defect image, the feature values ​​of each dimension are processed by Z score standardization, and the 3-times standard deviation principle is used to eliminate abnormal data for the standardized shape feature vector and size feature vector, and the feature vector after eliminating the abnormal value is aligned to obtain the second feature vector. For the feature values ​​in the second feature vector, the Pearson correlation coefficient is used to calculate the correlation between the shape feature and the size feature. If the correlation coefficient is positive, the two features are positively correlated and retained. If the correlation coefficient is negative, the two features are negatively correlated and retained. The feature values ​​are weighted by the absolute value of the correlation coefficient to obtain the third feature vector. The shape feature values ​​and size feature values ​​with assigned weights are extracted from the third feature vector, and the fused feature values ​​are obtained by weighted average calculation. The fused feature values ​​are weighted summed according to the calculated weight values, and the weighted summation result is mapped to the numerical range by maximum and minimum value normalization to obtain the fourth feature vector. For the normalized characteristic values ​​in the fourth eigenvector, principal component analysis is used to reduce the dimension of the characteristic values. The number of principal components is screened by the cumulative contribution rate threshold of 0.95. The comprehensive characteristic vector is constructed according to the principal component score to obtain the characteristic data of thermal deformation defects. The fusion process of the shape characteristic vector and the size characteristic vector of the thermal deformation defect needs to consider the distribution characteristics and correlation of the characteristic data. Z score standardization converts the original characteristic data into a standard normal distribution with a mean of 0 and a standard deviation of 1, so that the characteristics of different dimensions are comparable. Taking the arc radius in the shape characteristic vector as an example, the original values ​​are distributed between 10 and 30 mm, and the standardized values ​​are distributed between -2 and 2. For the determination of outliers, if the Z score of a measurement value exceeds 3 or is less than -3, it is regarded as abnormal data and is eliminated. The correlation analysis between features reflects the intrinsic connection between shape features and size features. In actual measurement, the arc radius of the bending defect is significantly positively correlated with the defect area, and the correlation coefficient reaches 0.82, indicating that the larger the arc radius, the larger the defect area. The torsion angle and the width of the defect area are negatively correlated, with a correlation coefficient of -0.65, indicating that the larger the torsion angle, the smaller the width of the defect area. The positive and negative signs of the correlation coefficient determine the direction of weight allocation during feature fusion. The feature fusion adopts the weighted average method, and the weight value is proportional to the absolute value of the correlation coefficient. For the arc radius feature with a correlation coefficient of 0.82, a weight of 0.35 is assigned; for the area feature with a correlation coefficient of 0.75, a weight of 0.32 is assigned; for the torsion angle feature with a correlation coefficient of 0.65, a weight of 0.28 is assigned; and for the remaining features, a weight of 0.05 is assigned. The fused feature values ​​after weighted summation are again mapped to the range of 0 to 1 through maximum and minimum value normalization. Principal component analysis reduces the dimension of the normalized fusion features and retains the principal components with a cumulative contribution rate of 95%. In practical applications, the contribution rates of the first three principal components are 65%, 20% and 10% respectively, and the cumulative contribution rate reaches 95%, so these three principal components are selected to construct the final comprehensive feature vector.The first principal component mainly reflects the overall size characteristics of the defect, the second principal component reflects the shape deformation characteristics, and the third principal component reflects the local detail characteristics. This dimensionality reduction process not only retains the key feature information, but also reduces data redundancy. In the process of feature fusion, the characteristics of different types of defects are different. Bending defects are mainly based on size features, supplemented by shape features; twisting defects are the opposite, with a higher weight for shape features; and shrinkage defects have similar weights for shape features and size features. Through feature fusion and dimensionality reduction, the comprehensive feature vector obtained can fully reflect all aspects of thermal deformation defects and provide reliable feature data for subsequent defect identification.

[0025] Step S105 , inputting the comprehensive feature vector of the thermal deformation defect into a preset thermal deformation defect recognition model for training, and learning the preset labeled samples to establish a discriminant function between the thermal deformation defect type and the comprehensive feature vector.

[0026] A training sample set is obtained through undersampling according to the comprehensive feature vector of thermal deformation defects, the number of bending defect samples, twisting defect samples and shrinkage defect samples in the training sample set is kept balanced, and a label matrix is ​​constructed for the training sample set; a nonlinear classifier is constructed for the label matrix using a support vector machine, and the radial basis kernel function parameters and the penalty coefficient are optimized by a grid search method to obtain an optimal parameter combination; a discriminant function is constructed using the optimal parameter combination, and the support vector is iteratively calculated by a sequence minimum optimization algorithm, and the discriminant function coefficient is obtained according to the interval maximization principle; a discriminant function value is calculated for the discriminant function coefficient, and if the discriminant function value is greater than the discriminant threshold, it is determined to be a first-class defect, and if the discriminant function value is less than the discriminant threshold, it is determined to be a second-class defect, and the defect recognition result is recorded through a confusion matrix to obtain a thermal deformation defect discriminant function.

[0027] Specifically, according to the comprehensive feature vector of thermal deformation defects and the preset labeled samples, the number of bending defect samples, twisting defect samples, and shrinkage defect samples is kept balanced through undersampling, the training sample set and the verification sample set are divided according to the ratio of 8:2, and the label matrix is ​​constructed for the defect labeling category information. The training subset is divided by 5-fold cross validation to obtain the fifth feature vector. For the fifth feature vector and the corresponding label data, a nonlinear classifier is constructed by a support vector machine. According to the radial basis kernel function parameter range of 0.1 to 10 and the penalty coefficient range of 1 to 100, the grid search method is used to optimize the parameters, and the classification accuracy is calculated from the verification subset to select the optimal parameter combination. The discriminant function is constructed from the optimal parameter combination, and the support vector is iteratively calculated by the sequence minimum optimization algorithm. The discriminant function coefficient is obtained according to the interval maximization principle. The iteration termination condition is set to the function value convergence less than 0.001 or the number of iterations reaches 1000 times to obtain the sixth feature vector. For the sixth eigenvector and the validation sample set, the discriminant function value of each defect sample is calculated, and the samples are classified by the discriminant threshold of 0. The classification accuracy, precision, and recall are calculated according to the predicted category and the true category. The confusion matrix is ​​used to record the recognition results of various types of defects to obtain the thermal deformation defect discriminant function. In the thermal deformation defect annotated samples, the number of defects of different types is often unbalanced. Taking the 1000 samples collected in actual production as an example, 500 are bending defects, 300 are twisting defects, and only 200 are shrinkage defects. The samples are kept balanced by undersampling, and 200 samples are randomly selected from bending defects and twisting defects, so that the number of samples of the three types of defects is 200. Among the 600 balanced samples, 480 training samples and 120 validation samples are obtained by dividing them in a ratio of 8:2. The core of the support vector machine is to map features to high-dimensional space through kernel functions. The parameter γ of the radial basis kernel function determines the complexity of the mapping space. The larger the γ value, the more complex the decision boundary. The grid search sets 10 value points in the range of 0.1 to 10, and searches for 10 value points of the penalty coefficient C from 1 to 100. In the actual optimization process, when γ is 0.5 and C is 10, the accuracy of the validation set reaches the highest value of 92%. The sequential minimum optimization algorithm solves the support vector by iteratively updating the Lagrange multiplier. In the classification of thermal deformation defects, support vectors are often distributed in the boundary area of ​​different types of defects. For example, at the boundary between bending defects and twisting defects, samples with an arc radius close to 45 degrees and slight torsion characteristics are likely to become support vectors. During the algorithm iteration process, the calculation is stopped when the difference in the objective function value of two consecutive iterations is less than 0.001, or the number of iterations exceeds 1000. The discriminant function directly reflects the defect category to which the sample belongs in practical applications. For a test sample, if its discriminant function value is greater than 0, it is classified as a positive class, that is, this type of defect, otherwise it is classified as a negative class, that is, other types of defects.The test on the validation sample set showed that the recognition accuracy of bending defects reached 94% and the recall rate was 91%; the accuracy of twisting defects was 89% and the recall rate was 87%; the accuracy of shrinkage defects was 90% and the recall rate was 88%. The confusion matrix further showed that the misclassification rate between bending defects and shrinkage defects was the lowest, at 3%, while the misclassification rate between twisting defects and bending defects was higher, reaching 8%, which was due to the overlap of some features between the two types of defects. During the entire process of model training and validation, 5-fold cross validation provided a stable performance evaluation. In each fold validation, the model was trained on a different training subset and validated on the remaining subset, and the average performance was finally taken as the evaluation result. This validation method effectively avoids overfitting, so that the trained discriminant function has good generalization ability.

[0028] Step S106, using the established discriminant function between the thermal deformation defect type and the comprehensive feature vector, extract the comprehensive feature vector from the input plastic lace thermal deformation defect image, obtain the thermal deformation defect recognition result according to the output result, and determine the defect type through the discriminant function of the thermal deformation defect recognition result, and the defect type includes bending, twisting and shrinkage.

[0029] For the shape feature vector and the size feature vector, the feature dimension corresponding to the cumulative contribution rate is obtained through principal component analysis, and a comprehensive feature vector is constructed by serial combination, and the data of the comprehensive feature vector is normalized; according to the comprehensive feature vector after data normalization, the preset radial basis kernel function parameters and penalty coefficient are used to calculate the distance from the sample to the classification hyperplane, and after conversion through the sigmoid function, the probability value of the defect type is obtained, and the discriminant value is calculated through the Boolean multi-class discriminant function. If the discriminant value exceeds the preset discriminant threshold, it is classified as the current defect type; for the calculated discriminant value, the maximum probability criterion is used to select the defect type, and if the discrimination result exceeds the preset confidence threshold, the thermal deformation defect type is determined.

[0030] Specifically, according to the shape feature vector and size feature vector extracted from the thermal deformation defect image, the feature dimension with a cumulative contribution rate of 95% is retained through principal component analysis, and a comprehensive feature vector is constructed by serial combination for the feature data after dimension reduction, and the combined feature vector is normalized to obtain the seventh feature vector. For the normalized feature data in the seventh feature vector, the distance from the sample to the classification hyperplane is calculated by the radial basis kernel function parameter 0.5 and the penalty coefficient 10 in the pre-trained support vector machine, and the defect type probability distribution is obtained by the sigmoid function conversion according to the distance value to obtain the eighth feature vector. The defect type probability value is extracted from the eighth feature vector, and the discriminant value of each type of defect is calculated by the Boolean multi-class discriminant function. The defect type is classified according to the preset discriminant threshold of 0.5. If the discriminant value exceeds the threshold, it is classified as the current defect type to obtain the ninth feature vector. For the discriminant value in the ninth feature vector, the maximum probability criterion is used to select the type with the largest probability value from bending defects, twisting defects, and shrinkage defects, and the recognition results are screened by the confidence threshold of 0.8. The thermal deformation defect type is determined according to the discrimination result to obtain the thermal deformation defect recognition result. After feature extraction, the thermal deformation defect image forms a high-dimensional feature vector, which includes two categories: shape features and size features. The shape features include 10 feature dimensions such as arc radius, torsion angle, and shrinkage ratio, and the size features include 8 feature dimensions such as length, width, and area. The principal component analysis reduces the dimension of these 18-dimensional features. The contribution rate of the first principal component is 65%, reflecting the overall size information of the defect, the contribution rate of the second principal component is 20%, reflecting the degree of shape deformation, and the contribution rate of the third principal component is 10%, representing the local detail features. Retaining these three principal components can cover 95% of the information. The pre-trained support vector machine uses the radial basis kernel function for feature mapping, and the kernel function parameter γ has a moderate feature mapping ability when it is 0.5. In practical applications, for a bending defect sample, its distance to the classification hyperplane is 2.5, and the probability value is 0.92 after conversion by the sigmoid function, indicating that the sample has a 92% probability of belonging to the bending defect type. In contrast, the probability of the same sample corresponding to the twisting defect is only 0.05, and the probability of the corresponding shrinkage defect is 0.03. The Boolean multi-class discriminant function realizes multi-class discrimination by constructing an inter-class distance matrix. In the actual classification, if the discriminant value of a bending defect sample is 0.85, which is higher than the preset threshold of 0.5, it is classified as a bending defect. At the same time, the discriminant value of the sample corresponding to the twist defect is 0.3, and the discriminant value corresponding to the shrinkage defect is 0.2, both of which do not exceed the discrimination threshold, further verifying the reliability of the classification results. The final defect type judgment adopts the maximum probability criterion, and a confidence threshold of 0.8 is set to ensure the reliability of the recognition result. Taking a certain sample as an example, the probabilities of the three defect types are 0.95 for bending defects, 0.03 for twisting defects, and 0.02 for shrinkage defects.The probability of bending defect is the largest and exceeds the confidence threshold of 0.8, so the sample is judged as a bending defect. For the sample with probability distribution of: bending defect 0.45, twisting defect 0.35, shrinkage defect 0.20, although the probability of bending defect is the largest, it does not exceed the confidence threshold. At this time, it is necessary to combine other feature information for further judgment. In the process of thermal deformation defect identification, the characteristics of different types of defects vary. Typical bending defects show obvious arc trajectories in the principal component space, twisting defects show spiral feature distribution, and shrinkage defects have the characteristics of centripetal aggregation. These features are fully reflected in the discriminant function, thereby achieving accurate defect type identification.

[0031] Step S107, obtaining the bending defect radius threshold, twisting defect torsion angle threshold and shrinkage defect shrinkage ratio threshold corresponding to the defect type. If the identified defect radius is less than the threshold, the torsion angle is greater than the threshold, and the shrinkage ratio is greater than the threshold, the plastic lace is judged to be unqualified in quality, otherwise it is judged to be qualified in quality.

[0032] The arc radius value of the bending defect, the torsion angle value of the twisting defect and the shrinkage ratio value of the shrinkage defect in the plastic lace sample are obtained, and the mean and standard deviation of the three types of defect characteristics are obtained through statistics of qualified samples; according to the mean and standard deviation of the defect characteristics, standardized processing is adopted to obtain standardized characteristic values; for the standardized characteristic values, the threshold interval boundaries are determined by adding and subtracting the standard deviation threshold from the mean, and the upper limit of the arc radius threshold, the lower limit of the torsion angle threshold and the lower limit of the shrinkage ratio threshold are obtained; if the arc radius value is less than the upper limit of the arc radius threshold and the torsion angle value is greater than the lower limit of the torsion angle threshold and the shrinkage ratio value is greater than the lower limit of the shrinkage ratio threshold, then it is determined that the quality of the plastic lace is unqualified, and the quality judgment mark is output by binarization processing.

[0033] Specifically, according to the identified thermal deformation defect type, the arc radius value of the bending defect, the torsion angle value of the twisting defect, and the shrinkage ratio value of the shrinkage defect are extracted, and the mean and standard deviation of the three types of defect characteristics are obtained through the statistics of qualified samples. The defect characteristic value is subtracted from the mean and then divided by the standard deviation to obtain the standardized characteristic value. For the standardized characteristic value, the upper limit of the arc radius threshold, the lower limit of the torsion angle threshold, and the lower limit of the shrinkage ratio threshold are calculated according to the principle of 3 times the standard deviation. The threshold interval boundary is determined by adding and subtracting 3 times the standard deviation from the mean, and the threshold range of each type of defect characteristic is recorded; the three types of defect threshold values ​​are extracted from them, and the degree of defect is determined by comparing the characteristic value with the threshold. If the arc radius value is less than the upper limit of the arc radius threshold and the torsion angle value is greater than the lower limit of the torsion angle threshold and the shrinkage ratio value is greater than the lower limit of the shrinkage ratio threshold, the plastic lace is judged to be unqualified in quality; based on the judgment result, the quality judgment mark is output by binary processing, and the mark value 0 indicates qualified quality and the mark value 1 indicates unqualified quality. The quality grade of the plastic lace is determined according to the mark value to obtain the quality judgment result of the thermal deformation defect. The quality judgment of the thermal deformation defect of plastic lace needs to consider multiple physical characteristic parameters. The radius of curvature of the bending defect in the qualified samples is usually between 15 mm and 25 mm, with a mean of 20 mm and a standard deviation of 1.7 mm. The torsion angle of the torsion defect is distributed between 20 degrees and 40 degrees, with a mean of 30 degrees and a standard deviation of 3.3 degrees. The shrinkage ratio of the shrinkage defect is between 0.85 and 0.95, with a mean of 0.9 and a standard deviation of 0.017. The standardization process converts these physical quantities to the same numerical scale. Taking the radius of curvature as an example, the radius value measured for a sample is 16 mm. After subtracting the mean of 20 mm and dividing by the standard deviation of 1.7 mm, the standardized value is -2.35. Similarly, the standardized value of the torsion angle of 35 degrees is 1.52, and the standardized value of the shrinkage ratio of 0.87 is -1.76. The standardized values ​​are convenient for unified comparison and threshold setting. The threshold interval is set based on the principle of 3 times standard deviation. The qualified lower limit of the arc radius is 15 mm, that is, the mean minus 3 times the standard deviation; the qualified upper limit of the torsion angle is 40 degrees, that is, the mean plus 3 times the standard deviation; the qualified upper limit of the shrinkage ratio is 0.95, that is, the mean plus 3 times the standard deviation. Samples beyond these threshold ranges are considered to be unqualified. In the actual judgment process, different types of defects often exist at the same time. Take a typical sample as an example. Its arc radius is 14 mm, which is less than the threshold of 15 mm; the torsion angle is 42 degrees, which is greater than the threshold of 40 degrees; the shrinkage ratio is 0.96, which is greater than the threshold of 0.95. These three parameters are beyond the qualified range at the same time, and the system judges it as unqualified quality and outputs the identification value 1. The arc radius of another sample is 18 mm, the torsion angle is 35 degrees, and the shrinkage ratio is 0.92. All three parameters are within the qualified range. The system outputs the identification value 0, indicating that the quality is qualified. The binary processing of the quality judgment results simplifies the subsequent processing flow. On the production line, an identification value of 0 directly triggers a release signal, while an identification value of 1 triggers an alarm signal.This judgment mechanism takes into account the combined effect of multiple defect characteristics, avoiding the one-sidedness that may be caused by single-feature judgment. For example, some samples may slightly exceed the standard in individual parameters, but other parameters perform well and may still be judged as qualified. This comprehensive judgment method meets the quality control needs in actual production.

[0034] The above only lists some preferred embodiments of the present invention, but the present invention is not limited thereto, and many improvements and changes can be made. As long as the improvements and changes are made on the basis of the basic principles of the present invention, they should be regarded as falling within the protection scope of the present invention.

Claims

1. A quality inspection method for plastic lace, characterized in that: The method comprises: The thermal deformation defect image of plastic lace is obtained, and image preprocessing is performed on the image, including denoising and enhancement, to obtain defect image data; relevant data of bending defects, twisting defects and shrinkage defects in the defect image data are extracted to obtain the defect area, and shape features are extracted from the relevant data, and the extracted shape features are constructed into feature vectors to obtain the shape feature vectors of the thermal deformation defects; the geometric size parameters of the defect area are measured, the total area of ​​the defect area and the area ratio relative to the entire image are calculated as relative area features, and the geometric size parameters of the defect area are combined into feature vectors to obtain the size feature vectors of the thermal deformation defects; the shape feature vectors and size feature vectors of the thermal deformation defects are fused to construct the comprehensive feature vectors of the thermal deformation defects; the comprehensive feature vectors of the thermal deformation defects are combined into a feature vector to obtain the size feature vectors of the thermal deformation defects; The eigenvector is input into a preset thermal deformation defect recognition model for training, and the preset labeled samples are learned to establish a discriminant function between the thermal deformation defect type and the comprehensive eigenvector; the discriminant function between the established thermal deformation defect type and the comprehensive eigenvector is used to extract the comprehensive eigenvector from the input plastic lace thermal deformation defect image, and the thermal deformation defect recognition result is obtained according to the output result. The thermal deformation defect recognition result is used to determine the defect type through the discriminant function, and the defect types include bending, twisting and shrinkage; the bending defect arc radius threshold, twisting defect torsion angle threshold and shrinkage defect shrinkage ratio threshold corresponding to the defect type are obtained. If the identified defect arc radius is less than the threshold and the torsion angle is greater than the threshold and the shrinkage ratio is greater than the threshold, the plastic lace is judged to be unqualified, otherwise the quality is judged to be qualified.

2. The method according to claim 1, characterized in that The method of obtaining the thermal deformation defect image of the plastic lace and performing image preprocessing on the image, wherein the preprocessing includes denoising and enhancement, and obtaining defect image data includes: For the original image of the thermal deformation defect of the plastic lace, the noise points in the original image are filtered out by median filtering, and the filtered image is processed by a bilateral filter to obtain a second image; Performing a block adaptive histogram equalization operation on the second image, and performing a nonlinear mapping transformation based on brightness features of a local area of ​​the image to obtain a third image; A Canny edge detection operator is used to extract the edge contour of the thermal deformation defect from the third image, the edge contour is processed by a morphological opening operation, and a grayscale threshold is used to segment the processed edge to obtain a fourth image; For the defect contour area segmented in the fourth image, the defect contour area is marked, and if the area of ​​the connected domain of the defect contour area is greater than the area threshold and the grayscale mean reaches a preset grayscale mean, the defect contour area is determined to be a target defect.

3. The method according to claim 1, characterized in that The method extracts relevant data of bending defects, twisting defects and shrinkage defects from the defect image data to obtain a defect area, extracts shape features from the relevant data, constructs the extracted shape features into a feature vector, and obtains a shape feature vector of the thermal deformation defect, including: The connected domain area of ​​the edge contour of the defect area is calculated, the edge contour curve is subjected to Hough circle transformation, the arc radius and arc angle of the curved defect area are obtained, and the shape features are extracted from the contour curve coordinate sequence to obtain the second image; For the marked area in the second image, using the image moment of inertia to calculate the main axis direction of the area; If the gray value of the marked area shows periodic changes in the main axis direction, it is determined to be a twist defect area; Obtaining the angle between the main axis direction of the twist defect area and the horizontal direction to obtain the torsion angle; Calculating the degree of twisting according to the fluctuation amplitude of the gray value of the twisted defect area in the main axis direction; Extracting shape features from the contour of the distortion defect area to obtain a third image; Extract the marked area from the third image, calculate the coordinates of the area centroid, and obtain the position of the shrinkage center point; The shrinkage direction is determined by the change in the distance from the boundary point of the extracted marked area to the centroid; The shrinkage ratio is calculated by taking the ratio of the area of ​​the shrinking region to the area of ​​the initial region; Extracting the shape features of the shrinkage defect according to the curvature change of the extracted contour points of the marked area to obtain a fourth image; For the shape features extracted from the fourth image, the minimum and maximum value normalization method is used to process the bending defect features, twisting defect features and shrinkage defect feature data respectively, and a thermal deformation defect shape feature vector is constructed according to the feature correlation weight to obtain the shape feature vector of the thermal deformation defect.

4. The method according to claim 1, characterized in that: The geometric size parameters of the defect area are measured, the total area of ​​the defect area and the area ratio relative to the entire image are calculated as relative area features, and the geometric size parameters of the defect area are combined into a feature vector to obtain a size feature vector of the thermal deformation defect, including: A Sobel operator is used to obtain an edge contour point set for the defect area, a projection length and a width are calculated according to the edge contour point set, an actual contour length is obtained by calculating the cumulative value of the contour point distance, and an actual width is obtained by the perpendicular distance between the contour point set and the projection axis; Extract the number of pixels occupied by the defective area, calculate the relative area ratio by the ratio of the number of pixels to the total number of pixels in the image, and use mean filtering to remove abnormal values ​​of the measurement data; The length, width and area values ​​of the defect area are normalized by minimum and maximum values. The feature weight is calculated by the Pearson correlation coefficient of the length, width and area values ​​of the defect area to obtain the size feature vector of the thermal deformation defect.

5. The method according to claim 1, characterized in that The shape feature vector and size feature vector of the thermal deformation defect are fused to construct a comprehensive feature vector of the thermal deformation defect, including: Obtaining a shape feature vector and a size feature vector in the defect image, performing Z-score standardization processing on the shape feature vector and the size feature vector, and eliminating abnormal data using the standard deviation principle to obtain a second feature vector; Calculate the Pearson correlation coefficient according to the characteristic values ​​in the second characteristic vector, if the correlation coefficient is a positive value, retain the positive correlation feature, if the correlation coefficient is a negative value, retain the negative correlation feature, and assign weights to the characteristic values ​​according to the absolute value of the correlation coefficient to obtain a third characteristic vector; Performing weighted average calculation on the feature values ​​to which weights have been assigned in the third feature vector, and performing numerical mapping on the weighted sum result by using maximum and minimum value normalization to obtain a fourth feature vector; A principal component analysis is performed on the normalized characteristic values ​​in the fourth eigenvector, the number of principal components is screened by a cumulative contribution rate threshold, and a comprehensive characteristic vector is constructed according to the principal component scores to obtain a comprehensive characteristic vector of the thermal deformation defect.

6. The method according to claim 1, characterized in that The method inputs the comprehensive feature vector of the thermal deformation defect into a preset thermal deformation defect recognition model for training, and learns the preset labeled samples to establish a discriminant function between the thermal deformation defect type and the comprehensive feature vector, including: A training sample set is obtained through undersampling processing according to the comprehensive feature vector of thermal deformation defects, the number of bending defect samples, twisting defect samples and shrinkage defect samples in the training sample set is kept balanced, and a label matrix is ​​constructed for the training sample set; A support vector machine is used to construct a nonlinear classifier for the label matrix, and the radial basis kernel function parameters and penalty coefficients are optimized by a grid search method to obtain an optimal parameter combination; The discriminant function is constructed by using the optimal parameter combination, the support vector is iteratively calculated by a sequence minimum optimization algorithm, and the discriminant function coefficient is obtained according to the interval maximization principle; The discriminant function value is calculated based on the discriminant function coefficient. If the discriminant function value is greater than the discriminant threshold, it is determined to be a first-class defect. If the discriminant function value is less than the discriminant threshold, it is determined to be a second-class defect. The defect recognition results are recorded through the confusion matrix to obtain the thermal deformation defect discriminant function.

7. The method according to claim 1, characterized in that The discriminant function between the established thermal deformation defect type and the comprehensive feature vector is used to extract the comprehensive feature vector from the input plastic lace thermal deformation defect image, and the thermal deformation defect recognition result is obtained according to the output result. The thermal deformation defect recognition result is used to determine the defect type through the discriminant function. The defect type includes bending, twisting and shrinkage, including: For the shape feature vector and the size feature vector, the feature dimension corresponding to the cumulative contribution rate is obtained through principal component analysis, a comprehensive feature vector is constructed by serial combination, and data normalization is performed on the comprehensive feature vector; According to the comprehensive feature vector after data normalization, the preset radial basis kernel function parameters and penalty coefficients are used to calculate the distance from the sample to the classification hyperplane. After conversion through the sigmoid function, the probability value of the defect type is obtained. The discriminant value is calculated through the Boolean multi-class discriminant function. If the discriminant value exceeds the preset discriminant threshold, it is classified as the current defect type. For the calculated discrimination value, the maximum probability criterion is used to select the defect type. If the discrimination result exceeds the preset confidence threshold, the thermal deformation defect type is determined.

8. The method according to claim 1, characterized in that The step of obtaining the bending defect radius threshold, twisting defect torsion angle threshold, and shrinkage defect shrinkage ratio threshold corresponding to the defect type, and if the identified defect radius is less than the threshold, the torsion angle is greater than the threshold, and the shrinkage ratio is greater than the threshold, the plastic lace is judged to be unqualified, otherwise, the quality is judged to be qualified, including: Obtain the arc radius value of the bending defect, the torsion angle value of the twisting defect, and the shrinkage ratio value of the shrinkage defect in the plastic lace sample, and obtain the mean and standard deviation of the three types of defect characteristics through statistics of qualified samples; According to the mean and standard deviation of the defect characteristics, a standardized characteristic value is obtained by performing a standardization process; For the standardized characteristic value, the threshold interval boundary is determined by adding and subtracting the standard deviation threshold from the mean value, and the upper limit of the arc radius threshold, the lower limit of the torsion angle threshold, and the lower limit of the contraction ratio threshold are obtained; If the arc radius value is less than the arc radius threshold upper limit, the torsion angle value is greater than the torsion angle threshold lower limit, and the shrinkage ratio value is greater than the shrinkage ratio threshold lower limit, the plastic lace is judged to be unqualified in quality, and a quality judgment mark is outputted by binarization processing.

Images