Intelligent monitoring method and system for PE pipe production

Abstract

The invention relates to the technical field of image processing, and particularly discloses an intelligent monitoring method and system for PE pipe production, and the method comprises the following steps: obtaining a wave pitch in a PE corrugated pipe image according to the collected PE corrugated pipe image; continuously intercepting a PE corrugated pipe image according to the length of 1-n wave pitches, and correspondingly obtaining a plurality of image block sets; taking the sum of the normalized value of the mean value of the similarity of the abnormal image block and the other image blocks in each image block set and the product of the corresponding first weight as the defect degree of the PE corrugated pipe; if the defect degree is larger than the threshold value, it is judged that the PE corrugated pipe has defects. According to the intelligent monitoring method and system for PE pipe production provided by the invention, the two image blocks can be ensured to be aligned, so that the calculation result is more accurate, and by integrating the results of the image block sets with multiple scales, the stability and robustness of detection are improved, and the occurrence of missing detection and false detection of defects in the PE pipe production process is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to an intelligent monitoring method and system for PE pipe production. Background Art

[0002] PE corrugated pipe, full name polyethylene corrugated pipe, is a plastic pipe extruded from polyethylene material with annular corrugations on both the inner and outer surfaces. PE corrugated pipe has the advantages of compact structure, good compensation effect, wear resistance, corrosion resistance, low temperature resistance, etc. Therefore, PE corrugated pipe is applied to fields such as municipal engineering, construction engineering, medical equipment, and instrumentation. The quality of the corrugated pipe is crucial for the stable operation of the equipment. However, the production of corrugated pipes requires multiple processes, and defects on the surface of the corrugated pipes can be caused by vibrations of processing equipment, misalignment of molds, and adhesion of waste chips, etc. Common defects include scratches, pits, small holes on the outer wall, etc. In order to prevent defective corrugated pipes from flowing out of the factory, it is necessary to inspect the corrugated pipes for defects. Common detection of surface defects of corrugated pipes mainly relies on visual inspection by human eyes, and the detection results are somewhat subjective, and the accuracy and efficiency fluctuate due to the state of the workers.

[0003] With the development of computer vision technology and image processing technology, the method of using machine vision to replace manual detection has achieved good results in various industrial fields. For example, in the Chinese patent application document with the application publication number CN114612434A, a method and system for detecting surface defects of corrugated pipes are disclosed. The method includes: obtaining a positive training set and a negative training set; constructing a corrugated pipe image training set based on the positive training set and the negative training set; training an image reconstruction network model based on the corrugated pipe image training set; obtaining an image of the corrugated pipe to be detected; inputting the image of the corrugated pipe to be detected into the image reconstruction network model to obtain a reconstructed image; and positioning the defects of the image of the corrugated pipe to be detected based on the reconstructed image and the image of the corrugated pipe to be detected. However, in the above solution, it is necessary to first obtain positive training images and negative training images according to the collected defect-free images, and train the network model based on this, and then input, process, and judge the image to be detected based on the network model. The calculation process of this method is relatively complex, and the operation time is relatively long. Since the corrugated pipe is in a conveying state, this method cannot timely judge the defects of the corrugated pipe, which may cause defective products to flow out of the factory.

[0004] In the prior art, a template matching algorithm is also used to detect defects in PE corrugated pipes. The process of the template matching algorithm is as follows: As the corrugated pipe moves, multiple images of the corrugated pipe are obtained as images to be detected, and the correlation between each image to be detected and a standard image is calculated respectively for matching. The image with a correlation less than the threshold is used as a defective image. Among them, the standard image is an image without defects. When calculating this template matching algorithm, it is necessary to ensure that the corresponding pixel points of the image to be detected and the standard image are aligned. Otherwise, when performing the matching calculation, an image to be detected without defects will be misjudged as having defects. However, due to the movement of the corrugated pipe, the corresponding pixel points of the multiple images to be detected captured and the standard image cannot be guaranteed to be aligned. Therefore, when using this algorithm to detect defects in the corrugated pipe, there will be a misdetection phenomenon, and the defects in the corrugated pipe cannot be accurately detected, thus affecting the monitoring effect during the production process of the corrugated pipe. Summary of the Invention

[0005] The present invention provides an intelligent monitoring method and system for PE pipe production, aiming to solve the technical problem in the prior art that due to the movement of the corrugated pipe, the corresponding pixel points of the image to be detected and the standard image cannot be guaranteed to be aligned, so there is a misdetection phenomenon, and the defects in the corrugated pipe cannot be accurately detected, thus affecting the defect monitoring effect during the production process of the corrugated pipe.

[0006] An intelligent monitoring method for PE pipe production according to the present invention includes the following steps: According to the collected PE corrugated pipe image, obtain the wave distance in the PE corrugated pipe image; Continuously intercept the PE corrugated pipe image with the length of 1 to n wave distances respectively, and correspondingly obtain multiple image block sets; where n ; Take the sum of the product of the normalized value of the mean similarity between the abnormal image blocks and the remaining image blocks in each image block set and the corresponding first weight as the defect degree of the PE corrugated pipe; among them, the abnormal image block is the image block with the smallest mean similarity with the remaining image blocks in the image block set; the first weight is the ratio of the standard deviation of the similarity between any two image blocks in any image block set to the sum of the standard deviations of the similarity between any two image blocks in all image block sets; If the defect degree is greater than the threshold, it is determined that the PE corrugated pipe has defects.

[0007] In the above solution, since the length dimensions and column pixel points of the image blocks within each image block set correspond to each other, any two image blocks can be guaranteed to be aligned, making the calculation results more accurate and facilitating the identification of abnormal image blocks. Moreover, through the multi-scale image block set, that is, by calculating the similarity of the image block set composed of image blocks with lengths that are different integer multiples of the wave distance to determine the defect degree of the corrugated pipe, and monitoring the production of the corrugated pipe based on the defect degree. The multi-scale analysis can integrate the results of the image block sets at multiple scales, thereby improving the stability and robustness of the detection and reducing the occurrence of missed detections and false detections of pipe defects during the pipe production process.

[0008] Preferably, obtaining the wave distance in the PE corrugated pipe image according to the collected PE corrugated pipe image includes: Establish a coordinate system with the column pixel points of the PE corrugated pipe image as the abscissa and the mean value of the gray values of the corresponding column pixel points as the ordinate, and use the maximum value of each mean value as the peak value; Take the sum of the product of the difference between any peak value and the abscissa of the previous adjacent peak value and the corresponding second weight as the wave distance in the PE corrugated pipe image; where the second weight is the ratio of the difference between any peak value and the minimum peak value to the sum of the differences between all peak values and the minimum peak value respectively.

[0009] In the above solution, by introducing the second weight to reflect the prominence degree of the peak value, the error when directly calculating the wave distance using the abscissa corresponding to the peak value is corrected, preventing the influence caused by the optical effect, and making the calculation of the wave distance in the PE corrugated pipe image more accurate.

[0010] Preferably, continuously intercept the PE corrugated pipe image with lengths of 1 to n wave distances, and intercept the PE corrugated pipe image within the range starting from the minimum abscissa corresponding to the peak value and ending at the maximum abscissa corresponding to the peak value.

[0011] In the above solution, by intercepting in this way, a complete corrugated image can be divided into one image block, which is convenient for subsequent similarity calculation between image blocks, and can prevent the non-complete corrugated images at both ends of the corrugated pipe image from being divided into one image block, resulting in a small similarity calculated between the complete image block and the non-complete image block and a large error.

[0012] Preferably, the similarity is the product of the normalized values corresponding to the similarity of the gray value vectors, the similarity of the gradient vectors, and the similarity of the local binary pattern eigenvalue vectors of any two image blocks; where the gray value vector is composed of the corresponding frequencies of the gray values of the pixel points in the image block sorted from small to large; the gradient vector is composed of the corresponding frequencies of the gradient values of the pixel points in the image block sorted from small to large; the local binary pattern eigenvalue vector is composed of the corresponding frequencies of the local binary pattern eigenvalues of the pixel points in the image block sorted from small to large.

[0013] In the above solution, the similarity between image blocks is calculated by combining the gray value, gradient value, and local binary pattern eigenvalue, so as to comprehensively judge abnormal image blocks and then find defects in the PE corrugated pipe.

[0014] Preferably, the similarity is represented by the cosine similarity of the corresponding vectors of two image blocks; Then the similarity is: ; In the formula, is the cosine similarity between the gray vector of image block and the gray vector of image block , is the cosine similarity between the gradient vector of image block and the gradient vector of image block , is the cosine similarity between the local binary pattern eigenvalue vector of image block and the local binary pattern eigenvalue vector of image block ; is a normalization function, , x is the independent variable, specifically , or .

[0015] Preferably, the gradient value of the pixel points in the image block is calculated by the Sobel operator.

[0016] Preferably, the acquisition process of the PE corrugated pipe image includes: Fix the image acquisition device in the radial direction of the PE corrugated pipe. As the PE corrugated pipe is conveyed along its axial direction, multiple PE corrugated pipe images are obtained to monitor the defects of each PE corrugated pipe image.

[0017] Preferably, the PE corrugated pipe images of two opposite semi-cylindrical circumferential walls of the PE corrugated pipe are collected simultaneously to monitor the defects of the entire circumferential wall of the PE corrugated pipe.

[0018] In the above solution, collecting the PE corrugated pipe images of two opposite semi-cylindrical circumferential walls of the PE corrugated pipe simultaneously can improve the efficiency of defect monitoring of the PE corrugated pipe and realize the defect monitoring of the entire circumferential wall of the PE corrugated pipe.

[0019] Preferably, the value range of the threshold is 0.5 to 0.7.

[0020] The present invention also provides an intelligent monitoring system for PE pipe production, including a memory and a processor. The processor executes the computer program stored in the memory to implement the intelligent monitoring method for PE pipe production described in any one of the above.

[0021] The beneficial effects are as follows: In the solution of the present invention, first, the wave distance in the PE corrugated pipe image is calculated through the gray values of the column pixel points in the PE corrugated pipe image. The PE corrugated pipe image is continuously intercepted at lengths that are different integer multiples of the wave distance to obtain multiple image block sets. Then, abnormal image blocks are obtained through the similarity of the image blocks in the image block sets. The defect degree of the PE corrugated pipe is obtained based on the similarity and the corresponding weights between the abnormal image blocks and the remaining image blocks in each image block set, so as to perform defect detection on the PE corrugated pipe and realize the intelligent monitoring of PE corrugated pipe production. Since the length dimensions and column pixel points of the image blocks in each image block set correspond to each other, any two image blocks can be guaranteed to be aligned, making the calculation results more accurate and facilitating the finding of abnormal image blocks. Moreover, through multi-scale image block sets, that is, the similarity of the image block sets composed of image blocks with lengths that are different integer multiples of the wave distance is used to calculate the defect degree of the corrugated pipe, and the production of the corrugated pipe is monitored through the defect degree. Multi-scale analysis can synthesize the results of multiple-scale image block sets, thereby improving the stability and robustness of detection and reducing the occurrence of missed detections and false detections of pipe defects during the pipe production process. Furthermore, in the solution of the present invention, there is no need to input a template image, and the image blocks are used as templates for each other to calculate similarity, nor is it necessary to construct a network model. Therefore, the calculation is relatively simple and the operation time is short. Description of the Drawings

[0022] By referring to the following detailed description with reference to the accompanying drawings, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become readily understood. In the drawings, several embodiments of the present invention are shown in an exemplary rather than restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts, where: Figure 1 is the step flow chart of the intelligent monitoring for PE pipe production according to the embodiment of the present invention; Figure 2 is the PE corrugated pipe image according to the embodiment of the present invention; Figure 3 is the coordinate system with the column pixel points of the PE corrugated pipe image as the abscissa and the average value of the corresponding gray values as the ordinate according to the embodiment of the present invention; Figure 4 is the structural block diagram of the intelligent monitoring system for PE pipe production according to the embodiment of the present invention. Detailed Embodiments

[0023] Embodiments of the present invention will be described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below by referring to the accompanying drawings are exemplary and are intended to explain the present invention and should not be construed as limiting the present invention.

[0024] As Figure 1 shown, according to a first aspect of the present invention, an intelligent monitoring method for PE pipe production is provided, including the following steps: S1. Obtain the wave distance in the PE corrugated pipe image according to the collected PE corrugated pipe image.

[0025] In this step, the image acquisition device is fixedly arranged in the radial direction of the PE corrugated pipe. As the PE corrugated pipe is conveyed along its axial direction, multiple PE corrugated pipe images are obtained to monitor the defects of each PE corrugated pipe image. The collected PE corrugated pipe images are as Figure 2 shown. In order to reduce the calculation amount, the PE corrugated pipe image is a black-and-white image. The image acquisition device can be a digital camera, a camera, etc.

[0026] In order to improve the efficiency of PE corrugated pipe defect monitoring, the PE corrugated pipe images of two opposite semi-cylindrical circumferential walls of the PE corrugated pipe can be collected simultaneously to monitor the defects of the entire circumferential wall of the PE corrugated pipe. Of course, the PE corrugated pipe images of one semi-cylindrical circumferential wall of the PE corrugated pipe can also be collected first to detect the defects of one semi-cylindrical circumferential wall of the PE corrugated pipe, and then the PE corrugated pipe is flipped 180 degrees, and the PE corrugated pipe images of the other semi-cylindrical circumferential wall of the PE corrugated pipe are collected to detect the defects of the other semi-cylindrical circumferential wall of the PE corrugated pipe.

[0027] Among them, the wave distance of the corrugated pipe refers to the distance between two adjacent wave crests in the corrugated pipe, which can be called the actual wave distance. The actual wave distance of the corrugated pipes of the same specification is fixed and unchanged. And the wave distance in the PE corrugated pipe image in the present invention refers to the distance corresponding to two adjacent wave crests in the PE corrugated pipe image. This wave distance is related to the distance between the image acquisition device and the PE corrugated pipe and the size of the image. The closer the image acquisition device is to the PE corrugated pipe, the larger this wave distance is, and the larger the size of the image is, the larger this wave distance is.

[0028] The PE corrugated pipe has a concave-convex corrugated structure, where the convex structure is called a wave crest and the concave structure is called a wave trough. Since the intensity of the reflected light on the wave crest and wave trough of the PE corrugated pipe is different, there will be bright and dark regions in the PE corrugated pipe image, as Figure 2As shown. Specifically, the brightness of the wave crest of the PE corrugated pipe in the image is greater than that of its wave trough in the image, that is, the wave crest corresponds to the bright area in the PE corrugated pipe image, and the wave trough corresponds to the dark area in the PE corrugated pipe image. And the brightness and darkness of the image are determined by the gray values of the pixel points in the image. Therefore, the positions of the wave crest and wave trough can be determined by using the gray values of each column of pixel points in the PE corrugated pipe image, and then the distance between adjacent wave crests can be calculated to obtain the wave distance in the PE corrugated pipe image. Specifically, step S1 includes the following steps: S11. Establish a coordinate system with the column pixel points of the PE corrugated pipe image as the abscissa and the mean value of the gray values of the corresponding column pixel points as the ordinate, and take the maximum value of each mean value as the peak value, as Figure 3 shown.

[0029] In this step, the PE corrugated pipe image is composed of multiple rows and columns of pixel points. Each column of pixel points corresponds to a position of the PE corrugated pipe along its axial direction. Therefore, the mean value of the gray values of each column of pixel points can reflect the brightness and darkness change of the PE corrugated pipe image along the axial direction. Among them, the larger the mean value of the gray values of each column of pixel points, the closer it is considered to be to the wave crest position of the PE corrugated pipe. On the contrary, the smaller the mean value of the gray values of each column of pixel points, the closer it is considered to be to the wave trough position of the corrugated pipe. And the maximum value of each mean value is considered to be at the wave crest position of the PE corrugated pipe. Therefore, the maximum value is used as the peak value.

[0030] S12. The sum of the product of the difference between any peak value and the abscissa corresponding to the previous adjacent peak value and the corresponding second weight is used as the wave distance in the PE corrugated pipe image; where the second weight is the ratio of the difference between any peak value and the minimum peak value to the sum of the differences between all peak values and the minimum peak value respectively.

[0031] In this step, ideally, the difference between the abscissas corresponding to two adjacent peak values can be calculated to represent the wave distance in the PE corrugated pipe image. However, in actual situations, not all peak values in the coordinate system are caused by the wave crests of the PE corrugated pipe. Sometimes, due to optical effects such as light reflection, some small peak values will appear in the coordinate system, such as Figure 3 the peak value B in.

[0032] Therefore, it is necessary to judge which peak values are caused by the wave crests of the PE corrugated pipe and which are caused by optical effects. According to the above, the peak values caused by optical effects (such as Figure 3 the peak value B in) are much smaller than the peak values caused by the wave crests (such as Figure 3The peak in (A), so the minimum peak (i.e., the minimum value of the peaks) can be considered to be caused by optical effects. And the larger the difference from the minimum peak, the more likely it is that the peak is caused by the wave crest of the PE corrugated pipe, such as Figure 3 For the peak A in, the smaller the difference from the minimum peak, the more likely it is that the peak is caused by optical effects. Therefore, the ratio of the difference between any peak and the minimum peak to the sum of the differences between all peaks and the minimum peak is used as the second weight to characterize the prominence of the peak, so as to correct the error when calculating the wave distance using the abscissa corresponding to the peak. Then the second weight corresponding to the th peak is: In the formula, is the th peak in the coordinate system, is the minimum peak in the coordinate system, is the number of peaks in the coordinate system.

[0033] Therefore, the wave distance W in the PE corrugated pipe image is: ; In the formula, is the second weight corresponding to the th peak, is the abscissa corresponding to the th peak, is the abscissa corresponding to the th peak, is the number of peaks in the coordinate system.

[0034] In this step, by introducing the second weight to reflect the prominence of the peak, the error when directly calculating the wave distance using the abscissa corresponding to the peak is corrected, and the influence caused by optical effects is prevented, making the calculation of the wave distance in the PE corrugated pipe image more accurate.

[0035] In some alternative embodiments, first set a peak preset value, discard the peaks smaller than the peak preset value, and retain the peaks larger than the peak preset value. Then, calculate the average value of the differences in abscissas corresponding to two adjacent peaks among the remaining peaks, and this average value is used as the wave distance in the PE corrugated pipe image. This is because the peaks caused by the wave crests are quite different from the peaks caused by optical effects. Therefore, the peaks caused by optical effects can be discarded by setting the peak preset value, and the remaining peaks are the peaks caused by the wave crests. Then, through subsequent calculations, the accurate wave distance in the PE corrugated pipe image can be obtained.

[0036] S2. Continuously intercept the PE corrugated pipe image in lengths of 1 to n wave distances respectively, and correspondingly obtain multiple image block sets.

[0037] Among them, n , and n is an integer, that is, the PE corrugated pipe image is intercepted at a length that is an integer multiple of the wave pitch. For example, if the length of a single wave pitch is set to j , then the length of 2 times the wave pitch is 2 j , and the length of n times the wave pitch is nj .

[0038] In this step, the PE corrugated pipe images are continuously intercepted at different integer multiples of the wave pitch length, and multiple image block sets can be obtained. Specifically, the PE corrugated pipe image is continuously intercepted at the length of 1 wave pitch to obtain an image block set composed of image blocks with a length of 1 wave pitch; the PE corrugated pipe image is continuously intercepted at 2 times the wave pitch length to obtain an image block set composed of image blocks with a length of 2 times the wave pitch; until the PE corrugated pipe image is continuously intercepted at n times the wave pitch length to obtain an image block set composed of image blocks with a length of n times the wave pitch. In this way, multiple image block sets respectively composed of image blocks with different integer multiples of the wave pitch length can be obtained.

[0039] This step is because the image block with a length of 1 wave pitch is relatively narrow, and the range of pixel points included in this image block is small. At this time, the noise of the image will cause a certain interference to the defect detection of the PE corrugated pipe. For example, the noise of the image includes current interference and light interference of the device when collecting the image. Therefore, the present invention adopts multi-scale image blocks to reduce the interference caused by noise to the defect monitoring of the PE corrugated pipe.

[0040] Among them, the starting point of the PE corrugated pipe image interception is the minimum abscissa corresponding to the peak value, and the ending point is the maximum abscissa corresponding to the peak value. By intercepting in this way, a complete corrugated image can be divided into one image block, which is convenient for subsequent similarity calculation between image blocks, and can prevent the non-complete corrugated images at both ends of the corrugated pipe image from being divided into one image block, resulting in a small similarity calculated between the complete image block and the non-complete image block, and a large error.

[0041] Of course, the starting point of the PE corrugated pipe image interception can also be the minimum abscissa corresponding to the valley value of the wave trough, and the ending point is the maximum abscissa corresponding to the valley value of the wave trough. Intercepting in this way can also divide a complete corrugated image into one image block.

[0042] S3. Take the sum of the product of the normalized value of the mean of the similarities between the abnormal image blocks and the remaining image blocks in each image block set and the corresponding first weight as the defect degree of the PE corrugated pipe.

[0043] In this step, the abnormal image block is the image block with the minimum average similarity to the remaining image blocks in the image block set. Since the sizes and pixel points of the image blocks in the same image block set correspond to each other, if the PE corrugated pipe has no defects, the similarities of the individual image blocks are relatively high. If the PE corrugated pipe has defects, the similarity of the defective image block to the remaining normal image blocks is relatively low, and the defective image block is the above-mentioned abnormal image block. An abnormal image block can be found in each image block set, and the sum of the products of each abnormal image block and the corresponding first weight is used as the defect degree of the PE corrugated pipe. Since the defect degree of the corrugated pipe is calculated by the similarity of the multi-scale image block set, that is, the image block set composed of image blocks with lengths that are different integer multiples of the wave distance, and the production of the corrugated pipe is monitored by the defect degree, multi-scale analysis can integrate the results of the image block sets of multiple scales, thereby improving the stability and robustness of the detection and reducing the occurrence of missed detections and false detections of pipe defects during the pipe production process.

[0044] Specifically, step S3 includes the following steps: S31. Obtain the abnormal image blocks in each image block set; The abnormal image block is the image block with the minimum average similarity to the remaining image blocks in the image block set. For example, the image block set includes four image blocks, namely: image block , image block , image block and image block . Since there is a similarity between any two image blocks, 6 similarities can be obtained for this image block set, namely , , , , and , where is the similarity between image block and image block , is the similarity between image block and image block , is the similarity between image block and image block , is the similarity between image block and image block , is the similarity between image block and image block , is the similarity between image block and image block . Calculate image block , image block , image block , image block The average similarity to other image blocks, and the image block with the smallest average is used as the abnormal image block.

[0045] Among them, the similarity is the product of the normalized values corresponding to the similarity of the gray value vectors, the similarity of the gradient vectors, and the similarity of the local binary pattern eigenvalue vectors of any two image blocks respectively.

[0046] Among them, the gray value vector is composed of the frequency corresponding to the gray values of the pixel points in the image block sorted from small to large. Those skilled in the art know that the value range of the gray value of the pixel point is from 0 to 255. Each pixel point in each image block corresponds to a gray value, and this gray value vector is composed of the frequency of occurrence of each gray value in the image block. For example, if the gray value 0 appears 1 time, the gray value 1 appears 5 times, the gray value 3 appears 0 times,..., and the gray value 255 appears 10 times, then this gray value vector is (1, 5, 0,..., 10).

[0047] The local binary pattern eigenvalue vector is composed of the frequency corresponding to the local binary pattern eigenvalues of the pixel points in the image block sorted from small to large. The basic idea of the local binary pattern is to compare each pixel point with its surrounding pixel points to obtain the local image structure. Specifically, a 3×3 square window is defined, with the central pixel point of the window as the threshold, and the gray values of its adjacent 8 neighborhood pixel points are compared with the gray value of the central pixel point of the window. If the gray value of the neighborhood pixel point is less than the gray value of the central pixel point, the value of this neighborhood pixel point is set to 0, otherwise, it is set to 1. In this way, after comparing the 8 pixel points in the neighborhood of a 3×3 window with the central pixel point, an 8-bit binary number will be generated. Converting this binary number to a decimal number can obtain 256 eigenvalues, that is, from 0 to 255. The local binary pattern eigenvalue vector is composed of the frequency of occurrence of each eigenvalue in the image block. The composition method is the same as that of the above gray value vector, and will not be elaborated here.

[0048] Similarly, the gradient vector is composed of the frequency corresponding to the gradient values of the pixel points in the image block sorted from small to large. Among them, the gradient value of the pixel point is used to describe the change degree of the gray value of the pixel point in the image block. The gradient value of the pixel point in the image block is calculated by the Sobel operator. Specifically, the Sobel operator uses two 3×3 matrix operators to convolve with the image block respectively to obtain the horizontal gradient value and the vertical gradient value, then the gradient value of the pixel point in the image block is the arithmetic square root of the sum of the squares of the corresponding horizontal and vertical gradient values.

[0049] It should be noted that when calculating the similarity between the gradient vectors of two image patches, it is necessary to ensure that the dimensions of the two gradient vectors are the same, that is, the number of gradient values in the two gradient vectors is the same. This can be achieved in the following way: First, merge the gradient values in the two gradient vectors, and then sort them in ascending order of the gradient values, keeping only one of the repeated gradient values; then, respectively obtain the frequencies of the merged gradient values corresponding to their appearances in the image patch. If the gradient value does not appear in the image patch, the frequency is 0; finally, calculate the similarity using the two newly obtained gradient vectors.

[0050] By calculating the similarity between image patches through the combination of gray values, gradient values, and local binary pattern feature values, abnormal image patches can be comprehensively judged, and then defects in the PE corrugated pipe can be found.

[0051] Take the image patch and the image patch to illustrate the calculation of the similarity between any two image patches. The similarity between the image patch and the image patch is: ; In the formula, is the cosine similarity between the gray vector of the image patch and the gray vector of the image patch , is the cosine similarity between the gradient vector of the image patch and the gradient vector of the image patch , is the cosine similarity between the local binary pattern feature value vector of the image patch and the local binary pattern feature value vector of the image patch ; is the normalization function, , x is the independent variable, specifically , or .

[0052] In this step, the cosine similarity is used to represent the similarity between two vectors. The cosine similarity can be applied to the operation of multi-dimensional vectors and can meet the calculation requirements. Moreover, the value range of the cosine similarity is from -1 to 1. When the value of the cosine similarity is 1, it indicates that the directions of the two vectors are the same, that is, the similarity between the two vectors is the highest. When the value of the cosine similarity is -1, it indicates that the directions of the two vectors are opposite, that is, the similarity between the two vectors is the lowest. The above normalization function makes the value range of the normalized cosine similarity calculated be from 0 to 1, which is convenient for subsequent calculations.

[0053] In some alternative embodiments, the similarity between two vectors can also be represented by the Pearson correlation coefficient, Euclidean distance, Manhattan distance, etc.

[0054] S32. Use the ratio of the standard deviation of the similarity between any two image patches in any image patch set to the sum of the standard deviations of the similarity between any two image patches in all image patch sets as the first weight.

[0055] A similarity can be obtained between any two image patches within each image patch set. The more discrete the distribution of the similarities within the image patch set, that is, the larger the standard deviation, the greater the difference between the image patches within the image patch set, and the more likely it is to detect defects on the PE corrugated pipe within the image patch set. Therefore, a greater weight should be given to this image patch set. On the contrary, the more concentrated the distribution of the similarities within the image patch set, that is, the smaller the standard deviation, the smaller the difference between the image patches within the image patch set, and the less likely it is to detect defects on the PE corrugated pipe within the image patch set. Therefore, a smaller weight should be given to this image patch set. Then, use the n first weight corresponding to the image patch set composed of image patches with a wave distance length of : ; In the formula, is the standard deviation of the similarities within the image patch set composed of image patches with a wave distance length of n , is the sum of the standard deviations of the similarities within all image patch sets, is the largest integer multiple of the wave distance when intercepting the PE corrugated pipe image. For example, is 5.

[0056] S33. Calculate the defect degree of the PE corrugated pipe.

[0057] The defect degree of the PE corrugated pipe is: ; In the formula, is the first weight corresponding to the image patch set composed of image patches with a wave distance length of n , is the average of the similarity between the abnormal image patches and the remaining image patches within the image patch set composed of image patches with a wave distance length of n , is the exponential function with the natural constant as the base, is the largest integer multiple of the wave distance when intercepting the PE corrugated pipe image.

[0058] S4. If the defect degree is greater than the threshold, it is determined that the PE corrugated pipe has defects.

[0059] In this step, when the defect degree is greater than the threshold value, it indicates that the defect degree of the PE corrugated pipe is relatively high, and it is determined that the PE corrugated pipe has a defect. The value range of the threshold is from 0.5 to 0.7. Preferably, the threshold is 0.6. Of course, the value range of the threshold can be adjusted according to the actual situation.

[0060] After it is determined that the PE corrugated pipe has a defect, the production equipment can be controlled to stop and alarm to prevent unqualified products from flowing out, and the operator can be notified in time to eliminate abnormal situations such as equipment failures, realizing intelligent monitoring of the production of PE corrugated pipes.

[0061] In the intelligent monitoring method for PE pipe production of the present invention, first, the wave distance in the PE corrugated pipe image is calculated through the gray values of the column pixel points in the PE corrugated pipe image. The PE corrugated pipe image is continuously intercepted at lengths that are different integer multiples of the wave distance to obtain multiple image block sets. Then, abnormal image blocks are obtained through the similarity of each image block in the image block set. According to the similarity and corresponding weights of the abnormal image blocks and the remaining image blocks in each image block set, the defect degree of the PE corrugated pipe is obtained, and thus the defect detection of the PE corrugated pipe can be carried out to realize the intelligent monitoring of the production of PE corrugated pipes.

[0062] Since the length dimensions of the image blocks and the column pixel points in each image block set correspond to each other, any two image blocks can be guaranteed to be aligned, making the calculation result more accurate and facilitating the finding of abnormal image blocks. And by calculating the similarity of the multi-scale image block sets, that is, the image block sets composed of image blocks with lengths that are different integer multiples of the wave distance, to obtain the defect degree of the corrugated pipe, and carrying out production monitoring of the corrugated pipe through the defect degree, multi-scale analysis can integrate the results of image block sets at multiple scales, thereby improving the stability and robustness of detection and reducing the occurrence of missed detection and false detection of pipe defects during the pipe production process. Moreover, in the solution of the present invention, there is no need to input a template image, and the similarity between each image block is calculated with each other as a template, and there is no need to construct a network model, so the calculation is relatively simple and the operation time is short.

[0063] As Figure 4 shown, according to the second aspect of the present invention, an intelligent monitoring system for PE pipe production is further provided. The system includes a memory and a processor, and the processor executes the computer program stored in the memory to implement the intelligent monitoring method for PE pipe production described in the first aspect of the present invention.

[0064] The system further includes a communication bus, a communication interface, and other components well known to those skilled in the art. Their settings and functions are known in the art, so they will not be described in detail here.

[0065] In the present invention, the aforementioned memory can be any tangible medium that contains or stores a program, which can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory (RRAM), a dynamic random access memory (DRAM), a static random access memory (SRAM), an enhanced dynamic random access memory (EDRAM), a high-bandwidth memory (HBM), a hybrid memory cube (HMC), etc., or any other medium that can be used to store the required information and can be accessed by an application program, module, or both. Any such computer storage medium can be part of the device or accessible or connectable to the device. Any application or module described in the present invention can be implemented by computer-readable / executable instructions stored or otherwise held by such a computer-readable medium.

[0066] In the description of this specification, the meaning of "a plurality of" is at least two, such as two, three, or more, etc., unless otherwise specifically and clearly defined.

[0067] Although this specification has shown and described multiple embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will think of many changes, alterations, and alternative ways without departing from the spirit and concept of the present invention. It should be understood that various alternatives to the embodiments of the present invention described herein can be adopted in the practice of the present invention.

Claims

1. An intelligent monitoring method for PE pipe production, characterized in that, It includes the following steps: Based on the collected PE corrugated pipe images, obtain the wave distance in the PE corrugated pipe images; Continuously intercept the PE corrugated pipe images at lengths of 1 to n wavelength pitches respectively, and correspondingly obtain multiple image block sets; among them, n ; Take the sum of the products of the normalized values of the means of the similarities between the abnormal image blocks and the remaining image blocks in each image block set and the corresponding first weight as the defect degree of the PE corrugated pipe; wherein, the abnormal image block is the image block with the smallest mean similarity to the remaining image blocks in the image block set; the first weight is the ratio of the standard deviation of the similarities between any two image blocks in any image block set to the sum of the standard deviations of the similarities between any two image blocks in all image block sets; If the defect degree is greater than the threshold value, it is determined that the PE corrugated pipe has defects.

2. The intelligent monitoring method for PE pipe production according to claim 1, characterized in that, The step of obtaining the wave distance in the PE corrugated pipe images based on the collected PE corrugated pipe images includes: Establish a coordinate system with the column pixel points of the PE corrugated pipe image as the abscissa and the mean of the gray values of the corresponding column pixel points as the ordinate, and take the maximum value of each mean as the peak value; Take the sum of the products of the difference between any peak value and the abscissa corresponding to the previous adjacent peak value and the corresponding second weight as the wave distance in the PE corrugated pipe image; wherein, the second weight is the ratio of the difference between any peak value and the minimum peak value to the sum of the differences between all peak values and the minimum peak value respectively.

3. The intelligent monitoring method for PE pipe production according to claim 2, wherein, The lengths of the wave pitches are respectively 1 to n continuously intercept the PE corrugated pipe image. Intercept the PE corrugated pipe image within the range starting from the minimum abscissa corresponding to the peak value and ending at the maximum abscissa corresponding to the peak value.

4. The intelligent monitoring method for PE pipe production according to claim 1, characterized in that, The similarity is the product of the normalized values corresponding to the similarities of the gray value vectors, gradient value vectors, and local binary pattern eigenvalue vectors of any two image blocks respectively; wherein, the gray value vector is composed of the corresponding frequencies of the gray values of the pixel points in the image block sorted from small to large; the gradient value vector is composed of the corresponding frequencies of the gradient values of the pixel points in the image block sorted from small to large; the local binary pattern eigenvalue vector is composed of the corresponding frequencies of the local binary pattern eigenvalues of the pixel points in the image block sorted from small to large.

5. The intelligent monitoring method for PE pipe production according to claim 4, characterized in that, The similarity is represented by the cosine similarity of the corresponding vectors of the two image blocks; Then the similarity is: ； Wherein, is the cosine similarity between the gray vector of image block and the gray vector of image block ; is the cosine similarity between the gradient vector of image block and the gradient vector of image block ; is the cosine similarity between the local binary pattern eigenvalue vector of image block and the local binary pattern eigenvalue vector of image block ; is a normalization function, , where x is the independent variable, specifically , or .

6. The intelligent monitoring method for PE pipe production according to claim 4, wherein The gradient values of the pixel points in the image block are calculated by the Sobel operator.

7. The intelligent monitoring method for PE pipe production according to claim 1, characterized in that, The process of collecting the PE corrugated pipe images includes: Fix the image acquisition device in the radial direction of the PE corrugated pipe, and as the PE corrugated pipe is conveyed along its axial direction, obtain multiple PE corrugated pipe images to monitor the defects of each PE corrugated pipe image.

8. The intelligent monitoring method for PE pipe production according to claim 1, characterized in that, Collect the PE corrugated pipe images of two opposite semi-cylindrical circumferential walls of the PE corrugated pipe simultaneously to monitor the defects of the entire circumferential wall of the PE corrugated pipe.

9. The intelligent monitoring method for PE pipe production according to claim 1, wherein The value range of the threshold is from 0.5 to 0.

7.

10. An intelligent monitoring system for PE pipe production, comprising a memory and a processor, characterized in that, The processor executes the computer program stored in the memory to implement the intelligent monitoring method for PE pipe production as described in any one of claims 1 to 9.

Images