A method for detecting sizing reed based on machine vision

Abstract

The present invention belongs to the technical field of machine vision detection, and relates to a method for detecting sizing yarn hitting a reed based on machine vision. The present invention provides a method for detecting sizing yarn hitting a reed based on machine vision, which can detect and track the phenomenon of warp yarn hitting a reed during the sizing process, and is a method for detecting sizing yarn hitting a reed with high accuracy, strong robustness, and wide variety adaptability. The complete image of the warp yarn passing through the reed teeth is captured by an image acquisition system that is independently built and can adapt to different types of sizing machines. First, the initial image is cropped to obtain the detection area, and then the image is smoothed by Gaussian filtering; finally, by improving the inter-frame difference method, expanding the inter-frame difference, adding the Canny edge algorithm and mathematical morphology operation, the detection and tracking of the warp yarn hitting the reed are realized; at the same time, the proposed algorithm can be applied to detection and recognition problems in other fields.

Description

Technical Field

[0001] The invention belongs to the technical field of machine vision detection, and relates to a sizing yarn collision reed detection method based on machine vision. Background Art

[0002] Sizing is an important preparatory process before weaving in the textile industry. Sizing gives the warp yarn the ability to resist complex external mechanical forces, improves the weavability of the warp yarn, and ensures the smooth progress of the weaving process. The warp yarn is broken by the telescopic reed to form yarn jamming, which is a common problem in the sizing process. If the reed jamming phenomenon is not discovered in time, the sizing machine shutdown rate will increase, the loss of warp yarn will increase, and defects will be generated in the subsequent weaving production, reducing the quality. At present, enterprises generally use manual inspection methods to deal with such problems, but manual inspection will increase the workload of workers, reduce production efficiency, and affect the benefits of enterprises. Therefore, textile enterprises are in urgent need of an accurate and efficient sizing reed detection method.

[0003] my country is accelerating the integration of new generation information technology and textile and garment industry, and promoting the development of textile and garment industry towards green, low-carbon, intelligent, digital and soft direction. As one of the advanced intelligent technologies, digital image technology is a new technology developed by the multidisciplinary integration of information technology and computer technology in the past two decades. With the development of computer technology, communication technology and Internet technology, modern technology can collect information data through different means, and the amount of information data is becoming larger and larger. With the development of image acquisition and pattern recognition technology, the application of machine vision quality control detection has become a trend for enterprises to implement timely response strategies, large-scale production and quality control; machine vision detection technology has the characteristics of non-contact, fast response and strong anti-interference ability on site, which provides new methods and means for automatic detection. Introducing machine vision technology into the detection of sizing yarn collision and reed yarn jamming can well meet its requirements for sensitivity and reliability. Sizing yarn is an important preparatory process before weaving in the textile industry. Sizing yarn gives the warp yarn the ability to resist external complex mechanical forces, improves the weavability of the warp yarn, and ensures the smooth progress of the weaving process. The warp yarn is broken by the telescopic reed to form a jamming yarn, which is a common problem in the sizing process. If the phenomenon of yarn collision and reed jam is not discovered in time, the downtime rate of the sizing machine will increase, the loss of warp yarn will increase, and defects will be produced in the subsequent weaving production, reducing the quality. At present, enterprises generally use manual inspection methods to deal with such problems, but manual inspection will increase the workload of workers, reduce production efficiency, and affect the benefits of enterprises. Therefore, textile enterprises urgently need a set of accurate and efficient sizing reed jam detection methods.

[0004] my country is accelerating the integration of new generation information technology and textile and garment industry, and promoting the development of textile and garment industry towards green, low-carbon, intelligent, digital and soft direction. As one of the advanced intelligent technologies, digital image technology is a new technology developed by the multidisciplinary integration of information technology and computer technology in the past two decades. With the development of computer technology, communication technology and Internet technology, modern technology can collect information data through different means, and the amount of information data is becoming larger and larger. With the development of image acquisition and pattern recognition technology, the application of machine vision quality control detection has become a trend for enterprises to implement timely response strategies, large-scale production and quality control; machine vision detection technology has the characteristics of non-contact, fast response and strong anti-interference ability on site, which provides new methods and means for automatic detection. Introducing machine vision technology into the detection of sizing yarn, reed and yarn can well meet its requirements for sensitivity and reliability. Summary of the invention

[0005] In order to overcome the problems of the prior art, the present invention aims to provide a sizing yarn collision reed detection method based on machine vision, which can detect the sizing yarn warp collision reed phenomenon in real time, and is a sizing yarn collision reed detection method with high accuracy, strong robustness and wide variety adaptability. At the same time, the proposed algorithm can identify, track and locate the sizing yarn collision reed position, and can be applied to detection and recognition problems in other fields, such as identification of moving targets, vehicle detection and other problems.

[0006] The technical solution of the present invention is as follows:

[0007] A method for detecting sizing reed collision based on machine vision comprises the following steps:

[0008] Step 1, collecting a wide range of images of the sized warp yarn at the reed teeth, the collected images include a wide range of images of the sized yarn passing through the reed teeth normally and the images of the sized yarn hitting the reed;

[0009] Step 2, cropping the collected image to obtain an image of the sizing yarn passing through the reed teeth;

[0010] Step 3, performing image smoothing processing on the image cropped in step 2 to reduce the image detail level;

[0011] Step 4, performing differential processing on the image in step 3 using an improved frame difference method to obtain a differential image;

[0012] Step 5, performing binary processing on the differential image obtained in step 4, and performing Canny edge algorithm and dilation mathematical morphological operation processing on the binary image;

[0013] Step 6, setting an area threshold, and using the area threshold as a sizing yarn hitting the reed judgment condition. If the contour area in step 5 is greater than the set area threshold, it is determined that the sizing yarn hits the reed. Otherwise, the sizing yarn passes through the reed teeth normally and does not hit the reed.

[0014] Step 7, calibrate the place where the sizing yarn hits the reed, so as to realize the identification, tracking and detection of the sizing yarn hitting the reed.

[0015] The specific steps of collecting the sizing image in step 1 are:

[0016] At different sizing machine speeds, collect as many types of images of the sizing yarn passing through the reed teeth as possible, using a common digital or fixed-focus camera for image acquisition. When collecting, the image quality should be guaranteed while ensuring that the largest possible width can be detected. The camera maintains a fixed distance from the sizing yarn, and the spatial resolution of the acquisition is recorded at the same time, so as to infer the actual distance characteristics.

[0017] The specific steps of image cropping in step 2 are:

[0018] Since the phenomenon of yarn collision and yarn jamming only occurs at the reed teeth of the sizing machine, in order to eliminate the interference caused by irrelevant areas outside the reed teeth and reduce the program running time, the collected images are cropped to retain only the reed tooth area. The cropped image can eliminate the interference caused by yarn jitter and obtain a complete image of the sizing yarn passing through the reed teeth.

[0019] The image smoothing in step 3 uses a Gaussian blur filter with the following specific features:

[0020] The convolution matrix composed of pixels with non-zero distribution is transformed with the original image, and the value of each pixel is the weighted average of the values ​​of the surrounding adjacent pixels. The value of the original pixel has the largest Gaussian distribution value, so it has the largest weight. As the adjacent pixels are farther away from the original pixel, their weights are getting smaller and smaller. This kind of blurring preserves the edge effect better than other balanced blur filters. The specific formula is:

[0021]

[0022] Where: f(x, y) is the Gaussian filter template coefficient, (x, y) is the point coordinate, and σ is the standard deviation of the normal distribution.

[0023] The specific characteristics of the improved frame difference method in steps 4 and 5 are:

[0024] The algorithm can extract the target of sizing yarn hitting the reed, and detect and track the sizing yarn hitting the reed in real time; the inter-frame difference method is to perform a difference operation on two consecutive frames of images in time, subtract the pixels corresponding to different frames, determine the absolute value of the grayscale difference, and obtain the difference image. The specific formula is:

[0025] D n (x, y) = |f n (x, y)-f n-1 (x, y)|#(2)

[0026] In formula (2), D n is the difference image, f n (x, y), f n-1 (x, y) are the nth frame and n-1th frame image respectively.

[0027] Steps 4 and 5 adopt the following improvements to address the problems of the traditional inter-frame difference method: 1) For the problem of missed detection of targets that change slowly, the frame difference is enlarged to deal with it. 2) The problem of target edge discontinuity in target detection is processed by adding the Canny edge algorithm, which can clearly and completely detect the edge of the yarn ball; in order to enhance the adaptability of the algorithm, the maximum inter-class variance method is used when selecting the dual threshold of the edge operator. 3) The warp yarns collide with the reed and overlap to produce holes, and the expansion mathematical morphology operation is used to fill the holes. The specific formula is:

[0028]

[0029] In formula (3), t is the enlarged frame difference, f n-t (x, y) is the nt frame image, D′ n To expand the differential image after the frame difference, N A is the total number of pixels in the area to be detected, λ is the light suppression coefficient, and A is the entire frame image to be detected.

[0030] Add the additional items in step 5.1 Indicates the change of the overall illumination in the image. When the illumination in the image changes greatly, the value of this item will increase, which can offset the sudden change of image difference caused by the illumination environment to a certain extent, thereby effectively suppressing the impact of sudden changes in the illumination environment on the detection results.

[0031] The main features of the sizing yarn hitting the reed detection and judgment in step 6 are:

[0032] When the warp yarn hits the reed, the yarn will gather. Therefore, after traversing the contour, the contour area threshold δ is used as the judgment condition for the warp yarn hitting the reed.

[0033]

[0034] In formula (4), R n is the target image detected last, that is, when the contour area>δ, it is determined that the warp yarn hits the reed and is surrounded by the yarn phenomenon; when the contour area≤δ, the warp yarn hits the reed and is surrounded by the yarn phenomenon does not occur.

[0035] The specific characteristics of the box calibration at the place where the sizing yarn hits the reed in step 7 are:

[0036] For the determination result of step 6, the bounding rectangle in the binary foreground image is used as the target image marking frame, and two adjacent sides of the bounding rectangle are parallel to the upper boundary and the lower boundary of the image respectively.

[0037] Beneficial effects of the present invention:

[0038] The present invention greatly reduces labor costs and is conducive to long-term maintenance; the improved frame difference method used in the present invention can be applied to the detection and tracking of sizing reeds of various sizing machines; it reduces warp yarn loss and cloth defects in post-weaving production, and improves the working efficiency and product quality of the sizing machine; the improved frame difference method based on machine vision proposed in the present invention can be applied to target detection and tracking problems in other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 The flowchart of the sizing reed detection method based on machine vision is as follows;

[0040] Figure 2 is a schematic diagram of an image acquisition device;

[0041] Figure 3 The figure is a schematic diagram of the sizing reed;

[0042] Figure 4 The algorithm flow chart for improving the frame difference method;

[0043] Figure 5 It is the difference diagram of sizing and reeding;

[0044] Figure 6 It is the morphological operation diagram of sizing and reeding;

[0045] Figure 7 This is the effect diagram of the sizing yarn collision reed detection. DETAILED DESCRIPTION

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the examples or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

[0048] The embodiment of the present invention provides a method for detecting sizing yarn collision with a reed based on machine vision, and its recognition flow chart is as follows: Figure 1In order to explain the specific implementation mode of the present invention in detail, the method of this embodiment includes the following steps:

[0049] Step 1, collect a wide range of images of sized warp yarns at the reed teeth, the collection device is as follows Figure 2 As shown, the collected images include a wide range of images of the sizing yarn passing through the reed teeth normally and the images of the sizing yarn hitting the reed;

[0050] In this step, the present invention collected 1840 sizing images from the laboratory as image data. Some sizing images can be referred to Figure 3 , the image size is 1920*1080 pixels;

[0051] Step 2, cropping the collected image;

[0052] Since the phenomenon of yarn collision and yarn jamming only occurs at the reed teeth of the sizing machine, in order to eliminate the interference caused by irrelevant areas outside the reed teeth and reduce the program running time, the image collected in step 1 is cropped to only retain the reed tooth area. The cropped image can eliminate the interference caused by yarn jitter and obtain a complete image of the sizing yarn passing through the reed teeth.

[0053] Step 3, smooth the image cropped in step 2 to reduce the level of image details; in this step, use a Gaussian blur filter to smooth the cropped image in step 2; the moving yarn in the detection process interferes with the automatic detection of the sizing yarn collision reed, and the Gaussian blur filter is used to smooth the cropped image to reduce the level of details. The value of the original pixel has the largest Gaussian distribution value, so it has the largest weight. The weight of the adjacent pixels becomes smaller as they are farther away from the original pixel. This blurring process retains the edge effect better than other balanced blur filters. The specific formula is:

[0054]

[0055] Where: f(x, y) is the Gaussian filter template coefficient, (x, y) is the point coordinate, and σ is the standard deviation of the normal distribution. Steps 4 and 5 establish an improved frame difference method target detection algorithm, and its flow chart is as follows: Figure 4 As shown in the figure, the algorithm can extract the target of the sizing yarn hitting the reed and perform real-time detection and tracking of the sizing yarn hitting the reed. The inter-frame difference method is to perform a difference operation on two consecutive frames of images in time, subtract the pixels corresponding to different frames, determine the absolute value of the grayscale difference, and obtain the difference image. The specific formula is:

[0056] D n (x, y) = |f n (x, y)-f n-1 (x, y)|#(2)

[0057] In formula (2), Dn is the difference image, f n (x, y), f n-1 (x, y) are the nth frame and n-1th frame image respectively. In this step, the following improvements are made to address the problems of the traditional frame difference method: 1) For the problem of missed detection of targets that change slowly, the method of expanding the frame difference is adopted to deal with it. 2) The problem of target edge discontinuity in target detection is processed by adding the Canny edge algorithm, which can clearly and completely detect the edge of the yarn ball; in order to enhance the adaptability of the algorithm, the maximum inter-class variance method is used when selecting the dual threshold of the edge operator. 3) The warp yarns collide with the reed and overlap to produce holes, and the expansion mathematical morphology operation is used to fill the holes. The specific formula is:

[0058]

[0059] In formula (3), t is the enlarged frame difference, f n-t (x, y) is the nt frame image, D′ n To expand the differential image after the frame difference, N A is the total number of pixels in the area to be detected, λ is the light suppression coefficient, and A is the entire frame image to be detected.

[0060] Add the additional items in step 5.1 Indicates the change of the overall illumination in the image. When the illumination in the image changes greatly, the value of this item will increase, which can offset the sudden change of image difference caused by the illumination environment to a certain extent, thereby effectively suppressing the impact of sudden changes in the illumination environment on the detection results.

[0061] Step 6: judging whether the yarn hits the reed or not based on the result processed in step 5;

[0062] In this step, the collision of the sizing yarn with the reed is determined; when the warp yarns collide with the reed, yarn agglomeration occurs.

[0063] Therefore, after traversing the contour, the contour area threshold δ is used as the judgment condition for the warp yarn to collide with the reed.

[0064]

[0065] In formula (4), R n is the target image detected last, that is, when the contour area>δ, it is determined that the warp yarn hits the reed and is surrounded by the yarn phenomenon; when the contour area≤δ, the warp yarn hits the reed and is surrounded by the yarn phenomenon does not occur.

[0066] Step 7 is to perform box calibration on the location where the sizing yarn hits the reed, so as to realize identification, tracking and detection of the sizing yarn hitting the reed.

[0067] In this step, for the determination result of step 6, the circumscribed rectangle is used as the target image marking frame to perform box calibration on the sizing yarn collision reed position; the adjacent two sides of the circumscribed rectangle are parallel to the upper boundary and the lower boundary of the image respectively.

[0068] In order to explain the specific effects of the present invention in detail, representative difference images, morphological operation images, and box-marked images can be referred to. Figure 5 , 6 、7.

[0069] Those skilled in the art should understand that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting sizing reed collision based on machine vision, characterized in that: The following steps are involved: Step 1, collecting images of the sized warp yarn at the reed teeth, the collected images include images of the sized warp yarn passing through the reed teeth normally and images of the sized warp yarn hitting the reed; Step 2, cropping the collected image to obtain an image of the sizing yarn passing through the reed teeth; The specific steps of image cropping in step 2 are: cropping the acquired image to retain only the reed tooth area; while eliminating the interference caused by yarn jitter, the cropped image obtains a complete image of the sizing yarn passing through the reed tooth; Step 3, performing image smoothing processing on the image cropped in step 2 to reduce the image detail level; The image smoothing in step 3 uses a Gaussian blur filter: The convolution matrix composed of pixels with non-zero distribution is transformed with the original image. The value of each pixel is the weighted average of the values ​​of the surrounding adjacent pixels. The value of the original pixel has the largest Gaussian distribution value, so it has the largest weight. As the adjacent pixels are farther away from the original pixel, their weights are getting smaller and smaller. The specific formula of the Gaussian blur filter is: Where: f(x,y) is the Gaussian filter template coefficient, (x,y) is the point coordinate, and σ is the standard deviation of the normal distribution; Step 4, performing differential processing on the image in step 3 using an improved frame difference method to obtain a differential image; Step 5, performing binary processing on the differential image obtained in step 4, and performing Canny edge algorithm and dilation mathematical morphological operation processing on the binary image; The specific features of the improved frame difference method in steps 4 and 5 are: 1) Processing by enlarging the frame difference; 2) Add the Canny edge algorithm for processing; when selecting the double threshold of the edge operator, the maximum inter-class variance method is used; 3) The warp yarns collide with the reed yarns and overlap to produce cavities, which are filled by the expansion mathematical morphology operation; the specific formula is: In formula (2), t is the enlarged frame difference, f n-t (x, y) is the nt-th frame image, D′ n To expand the differential image after the frame difference, N A is the total number of pixels in the area to be detected, λ is the light suppression coefficient, and A is the entire frame image to be detected; Step 6, setting an area threshold, and using the area threshold as a criterion for determining whether the sizing yarn hits the reed; In formula (3), R n is the target image detected last, that is, when the contour area>δ, it is determined that the warp yarn hits the reed and is surrounded by the yarn; when the contour area≤δ, the warp yarn hits the reed and is surrounded by the yarn; Area is the contour area; Step 7, calibrate the place where the sizing yarn hits the reed, so as to realize the identification, tracking and detection of the sizing yarn hitting the reed.

2. A method for detecting sizing reed based on machine vision as claimed in claim 1, characterized in that: In step 5, add the additional item It shows the change of overall illumination in the image.

3. A method for detecting sizing reed based on machine vision as claimed in claim 1 or 2, characterized in that: The specific steps of collecting images in step 1 are: collecting different types of images of sizing yarn passing through reed teeth at different sizing machine speeds, using an ordinary digital camera or a fixed-focus camera to collect images; the camera maintains a fixed distance from the sizing yarn, and records the spatial resolution of the collection at the same time, so as to infer the actual distance characteristics.

Images