Silk yarn defect detection method in production of spandex textile machine

Abstract

The invention belongs to the technical field of machine vision detection, and particularly relates to a silk thread defect detection method in spandex textile machine production, four light sources are arranged on each spandex textile machine, an AGV trolley is used as an inspection carrier of a collaborative robot, and three cameras with different focal lengths are installed on a mechanical arm of the collaborative robot. The first camera, the second camera and the third camera are arranged from small to large according to focal lengths; the first camera shoots images of the two groups of upper and lower compression rollers, and the system performs reverse winding defect detection on the images; the first camera shoots swing images of the silk thread on the first winding shaft and the second winding shaft, the second camera shoots swing images of the silk thread on the third winding shaft and the fourth winding shaft, and the system carries out swing anomaly detection on the swing images; and the third camera shoots an image of the silk thread between the first silk guide roller and the second silk guide roller, and the system performs silk skipping defect detection on the image. According to the invention, full-automatic detection of silk yarn defects in the spandex production process is realized, and the detection efficiency and accuracy are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of machine vision detection, and in particular relates to a method for detecting thread defects in spandex textile machine production. Background Art

[0002] At present, the main method of spandex defect inspection is manual inspection. Manual inspection method: Use a handheld flashlight to illuminate the spandex thread from the side of the spandex spinning machine, and judge whether the spandex thread is currently qualified by the presence or absence of the thread and the abnormal swing amplitude; then go to the next spandex spinning machine.

[0003] Each workshop has 12 textile lines, each line has 26 textile machines, a total of 312 textile machines; each line is equipped with a person who conducts continuous patrol inspections, which is a huge challenge to the workers' physical fitness. Long-term observation with a flashlight has a huge impact on the workers' vision health.

[0004] Common yarn defects in spandex textile machine production include broken yarn, doubling, skipped yarn and backwinding. Broken yarn: Broken yarn will cause shaft winding, which is manifested as one less yarn in the swing amplitude. Doubling: Doubling means one yarn is missing, two yarns are doubling into one, which is manifested as one less yarn in the swing amplitude. Skipped yarn: The overall number of yarns is the same, but in two adjacent bundles of yarn, one or more yarns jump from bundle a to bundle b, resulting in the overall number being the same, but bundle a has less yarn than bundle b. Backwinding: The yarn is not wound around the winding shaft, but is wound around the pressure roller by mistake. Summary of the invention

[0005] The purpose of the present invention is to propose a method for detecting thread defects in spandex textile machine production in view of the low efficiency of manual detection, the deviation caused by personal subjective judgment, and the missed detection caused by visual fatigue, so as to realize the fully automated detection of thread defects in the spandex production process and significantly improve the detection efficiency and accuracy.

[0006] To achieve the above purpose, the technical solution adopted is:

[0007] A method for detecting defects in silk threads in the production of a spandex textile machine, wherein each spandex textile machine comprises a first winding shaft, a second winding shaft, a third winding shaft, a fourth winding shaft, a first wire guide roller, a second wire guide roller and a third wire guide roller, and upper and lower pressure rollers are arranged between the first winding shaft and the second winding shaft, and between the third winding shaft and the fourth winding shaft, and a plurality of silk threads conveyed by the first wire guide roller are evenly wound on the first winding shaft, the second winding shaft, the third winding shaft and the fourth winding shaft via the second wire guide roller and the third wire guide roller; the method comprises:

[0008] Four light sources are arranged on each spandex textile machine, and an AGV is used as the inspection carrier of the collaborative robot. Three cameras with different focal lengths are installed on the robotic arm of the collaborative robot, which are the first camera, the second camera, and the third camera in ascending order of focal length.

[0009] The first camera takes images of two sets of upper and lower pressure rollers, and the system performs reverse winding defect detection on the images;

[0010] The first camera captures the swing images of the silk thread on the first winding shaft and the second winding shaft, and the second camera captures the swing images of the silk thread on the third winding shaft and the fourth winding shaft, and the system performs swing abnormality detection on the swing images;

[0011] The third camera captures an image of the wire between the first wire guide roller and the second wire guide roller, and the system performs wire skipping defect detection on the image.

[0012] According to the method for detecting silk thread defects in spandex textile machine production of the present invention, further, four light sources are installed on the side of the textile machine, and each light source is located above a winding shaft, for illuminating the entire row of silk threads conveyed to the winding shaft, and the light source is emitted from the transmitting end to the fixed end face of the winding shaft; the focal length of the first camera is 8 mm, the focal length of the second camera is 12 mm, and the focal length of the third camera is 25 mm.

[0013] According to the method for detecting thread defects in spandex textile machine production of the present invention, further, the emitting end of the light source is a circular emitting port with a diameter of 55mm-100mm, and light is focused by a plano-concave mirror to form a bright band.

[0014] According to the method for detecting yarn defects in spandex textile machine production of the present invention, further, the first camera captures the image of each set of upper and lower pressure rollers, and the system performs reverse winding defect detection on the image, comprising:

[0015] The AGV with the collaborative robot moves to the aisle in front of the winding shaft. At this time, the first camera is 1700mm-1800mm away from the ground, and the first camera is used to capture the image of the pressure roller.

[0016] Perform reverse winding defect detection on the dark roller in the image and extract the reverse winding defect using background difference method;

[0017] In the detection of the roller reverse winding defect with bright image, the background difference method is also first used to obtain the pending reverse winding defect area. Since the brightness of this roller is high and if there is interference from shaft dirt, further processing is required to distinguish true and false defects: first, the roller is extracted from the image using threshold segmentation. When the pending reverse winding defect is surrounded by a black background, it is determined to be a true reverse winding defect. When there is a certain gray value around the pending reverse winding defect, it is determined to be a false reverse winding defect.

[0018] Finally, the reverse winding defect occurs at the upper pressure roller and / or the lower pressure roller.

[0019] According to the method for detecting thread defects in spandex textile machine production of the present invention, further, the first camera captures a swing image of the thread on the first winding shaft and the second winding shaft, and the system performs swing abnormality detection on the swing image, comprising:

[0020] The AGV with the collaborative robot moves to the aisle in front of the winding shaft. At this time, the first camera is 1500mm-1700mm from the ground and is located to the right of the third wire guide roller. The first camera is used to capture the swing image.

[0021] Perform Gaussian filtering on the acquired swing image;

[0022] A swing detection model is constructed using a neural network, and the model is trained. The image to be detected after Gaussian processing is input into the trained swing detection model for swing detection to obtain the detection result, i.e., the position, width and height of the silk thread swing in the image;

[0023] To judge whether the thread swing is abnormal, specifically, the average thread swing width is obtained according to the number of threads, and then the thread swing width of each thread is compared with the average value to obtain the thread swing exceeding the average value as abnormal swing.

[0024] According to the method for detecting silk thread defects in spandex textile machine production of the present invention, further, the second camera captures the swing images of the silk thread on the third winding shaft and the fourth winding shaft, and when the system performs swing abnormality detection on the swing images, the second camera is 1700mm-1800mm from the ground and is located to the upper right of the third wire guide roller.

[0025] According to the method for detecting thread defects in spandex textile machine production of the present invention, further, the third camera captures an image of the thread between the first guide roller and the second guide roller, and the system performs wire skipping defect detection on the image, comprising:

[0026] The AGV with the collaborative robot moves to the aisle in front of the winding shaft. At this time, the third camera is 1500mm-1700mm from the ground and is located between the first wire guide roller and the second wire guide roller. The third camera is used to capture the wire jump image.

[0027] The collected skipped wire image is rotated using a rotation matrix so that the wire is in a vertical state in the image;

[0028] After rotating the image, threshold segmentation is performed in the ROI area to separate the second godet fixed end area from the background;

[0029] Calculate the eigenvalue of the row index row2 at the lower right corner of the fixed end area, further locate the detection area of ​​the thread based on the eigenvalue of row2, analyze the area again, and calculate the corresponding eigenvalue based on the camera position; generate the final detection area based on the above eigenvalues;

[0030] A wire skipping defect detection model is constructed using a neural network, and the model is trained. The processed images to be detected are input into the trained wire skipping defect detection model for wire skipping defect detection to obtain the detection results.

[0031] The beneficial effects achieved by adopting the above technical solution are:

[0032] The mobile visual inspection platform (AGV trolley + collaborative robot + visual acquisition system) of the present invention dynamically monitors each textile machine without manual real-time observation. The system can make independent judgments and send out alarms. It is a new detection system that integrates visual inspection technology and robotics technology. This system can adapt to larger-scale work sites, while taking into account the detection of single machines with insufficient space and aisles. In short, the present invention can promptly detect problems that occur in the production process of textile machines, such as backwinding, abnormal swing amplitude (broken wire defects) and wire skipping, improve detection efficiency and accuracy, save labor costs, and can help enterprises quickly locate the source of defects through defect information statistics, make quality improvements in a timely manner, and avoid producing a large number of defective products and causing cost waste.

[0033] The present invention uses a light source to illuminate the silk thread, so that the camera can accurately capture the swing amplitude, rewinding and wire skipping of the silk thread. For such a white translucent target object, the problem of not being able to capture the target object is solved by lighting from the side. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings of the embodiments of the present invention, wherein the drawings are only used to illustrate some embodiments of the present invention, but not to limit all embodiments of the present invention thereto.

[0035] Figure 1 It is a schematic flow chart of a method for detecting thread defects in spandex textile machine production according to an embodiment of the present invention;

[0036] Figure 2 is a front view of the arrangement of the godet roller, light source and camera of an embodiment of the present invention;

[0037] Figure 3 is a top view of the arrangement of the godet roller, light source and camera according to an embodiment of the present invention;

[0038] Figure 4 is a schematic diagram of the left upper and lower pressure rollers photographed by the first camera of an embodiment of the present invention;

[0039] Figure 5 yes Figure 4 Schematic diagram of the reverse winding defect captured by the first camera;

[0040] Figure 6 is a schematic diagram of the upper and lower pressure rollers on the right side photographed by the first camera in an embodiment of the present invention;

[0041] Figure 7 is a schematic diagram of the swing amplitude of the silk thread on the first winding shaft and the second winding shaft photographed by the first camera according to an embodiment of the present invention;

[0042] Figure 8 yes Figure 7 Schematic diagram of the swing anomaly captured by the first camera;

[0043] Fig. 9 is a schematic diagram of the swing of the silk thread on the third winding shaft and the fourth winding shaft photographed by the second camera of the embodiment of the present invention;

[0044] Fig.10 is a front view of the silk thread between the first wire guide roller and the second wire guide roller taken by the third camera of the embodiment of the present invention;

[0045] Fig.11 is a top view of the wire between the first wire guide roller and the second wire guide roller taken by the third camera in the embodiment of the present invention;

[0046] Fig.12 yes Fig.10 and Fig.11 Schematic diagram of wire skipping defect taken by the third camera. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings of specific embodiments of the present invention to clearly and completely describe the exemplary scheme of the embodiment of the present invention. Unless otherwise defined, the technical terms or scientific terms used in the present invention should be the common meanings understood by people with ordinary skills in the field.

[0048] The present embodiment discloses a method for detecting silk thread defects in the production of a spandex textile machine. Each spandex textile machine comprises a first winding shaft, a second winding shaft, a third winding shaft, a fourth winding shaft, a first wire guide roller, a second wire guide roller and a third wire guide roller. The first winding shaft and the second winding shaft are located below the second wire guide roller, and the third winding shaft and the fourth winding shaft are located below the third wire guide roller. Upper and lower pressure rollers are arranged between the first winding shaft and the second winding shaft, and between the third winding shaft and the fourth winding shaft. The pressure rollers are used to ensure the forming of the silk cake. When working normally, the silk thread will not be wound on the pressure rollers. When a reverse winding defect occurs, the silk thread is not wound on the winding shaft, but on the pressure roller. The 80 silk threads delivered by the first wire guide roller are passed through the second wire guide roller and the third wire guide roller, and 20 threads are respectively wound on the first winding shaft, the second winding shaft, the third winding shaft and the fourth winding shaft. As Figure 1 As shown, the method comprises the following steps:

[0049] Step S101: Figure 2 and Figure 3As shown, four light sources are arranged on each spandex textile machine, and an AGV car is used as the inspection carrier of the collaborative robot. Three cameras with different focal lengths are installed on the robotic arm of the collaborative robot, which are the first camera, the second camera and the third camera in order of focal length from small to large. The robotic arm carries three cameras to shoot five points to detect different defects.

[0050] The AGV is used as a carrier mechanism for inspection, and can be a track-guided or trackless guided type. The collaborative robot is an industrial collaborative robot.

[0051] Step S102: The first camera captures images of the two groups of upper and lower pressure rollers, and the system performs reverse winding defect detection on the images.

[0052] Step S103: the first camera captures the swing images of the silk thread on the first winding shaft and the second winding shaft, the second camera captures the swing images of the silk thread on the third winding shaft and the fourth winding shaft, and the system performs swing abnormality detection on the swing images.

[0053] Step S104: The third camera captures an image of the wire between the first wire guide roller and the second wire guide roller, and the system performs wire jump defect detection on the image.

[0054] Four light sources are installed on the side of the textile machine, and each light source is located above a winding shaft to illuminate the entire row of silk threads conveyed to the winding shaft. The light source is emitted from the emitting end to the fixed end face of the winding shaft. The emitting end of the light source is a circular emitting port with a diameter of 55mm-100mm. The light is concentrated by a plano-concave mirror to form a light band with a width of 70mm-120mm and a length of 1500mm-1800mm. The light source lamp beads are LED lamp beads with a full load of 350-550 lumens. The diameter of the plano-concave mirror is 60mm-100mm; the focal length of the first camera is 8mm, the focal length of the second camera is 12mm, and the focal length of the third camera is 25mm.

[0055] The camera is a 500W-2500W black and white high-definition global camera with a frame rate of 4.5 frames, a dynamic range of 63dB, and a signal-to-noise ratio of 36dB. The lens has a 500W-2500W resolution, an 8mm-25mm lens, a field of view of D: 20.32°, H: 16.77°, V: 11.24°, and a flange back focus of 17.526mm.

[0056] like Figure 4 and Figure 5 As shown, point 1: the first camera captures images of the upper and lower pressure rollers on the left side.

[0057] Step 1. The AGV carrying the collaborative robot moves to the aisle in front of the winding shaft. At this time, the first camera is 1700mm-1800mm from the ground and is located between the second wire guide roller and the third wire guide roller. The first camera is used to capture images of the two pressure rollers on the left.

[0058] Step 2: According to the position of the roller in the image, two ROI regions are predefined, and the image is cropped according to the ROI regions. Gaussian filtering is performed on the cropped image to eliminate some noise interference.

[0059] Step 3: Due to the lighting in the workshop and the camera angle, there are differences in the imaging of the two rollers. The upper roller is very bright and will reflect light, while the lower roller hardly reflects light. The lower roller has no interference. First, the reverse winding defect detection is performed on the lower roller in the image, and the reverse winding defect is extracted using the background difference method: the roller is mean filtered to obtain the current image, and then the current image is subtracted from the background image to obtain the possible defect area, and then the reverse winding defect is obtained through feature screening (the average grayscale of the area can be used for screening within a certain range of values).

[0060] Step 4: When detecting the reverse winding defect of the upper roller in the image, the background difference method is also used to obtain the pending reverse winding defect area (rectangle). Due to the shooting angle and workshop lighting, if there is dirt on the roller, the image on the upper roller will form a fuzzy area, which affects the detection of the real reverse winding defect. Further processing is required to distinguish between true and false defects: first extract the upper roller. Since the roller is reflective, it can be easily extracted in the image using threshold segmentation; when the pending reverse winding defect is surrounded by a black background, it is determined to be a real reverse winding defect. When there is a certain gray value around the pending reverse winding defect, it is determined to be a false reverse winding defect. The threshold segmentation formula is:

[0061]

[0062] Where T represents the preset segmentation threshold, g(x) represents the binary image, and f(x,y) represents the grayscale value of the pixel.

[0063] Step 5: Through the extraction of the reverse winding defect in steps 3 and 4, it is finally determined that the reverse winding defect occurs in the upper pressure roller and / or the lower pressure roller.

[0064] like Figure 6 As shown, point 2: the first camera takes images of the upper and lower pressure rollers on the right. The detection method is the same as point 1 above, except that the first camera is 1700mm-1800mm from the ground and is located above the third wire guide roller. The first camera is used to take images of the two pressure rollers on the right.

[0065] like Figure 7 and Figure 8 As shown, point three: the first camera captures the swing image of the silk thread on the first winding shaft and the second winding shaft.

[0066] Step 1: The AGV carrying the collaborative robot moves to the aisle in front of the winding shaft. At this time, the first camera is 1500mm-1700mm away and is located to the right of the third wire guide roller. The first camera is used to capture images.

[0067] Step 2: Perform Gaussian filtering on the acquired swing image to eliminate noise interference. Use discrete Gaussian to smooth the image. The discrete approximate formula of the Gaussian function is as follows:

[0068]

[0069] Where x represents a random variable and σ represents the variance.

[0070] Step 3: predefine the detection ROI, cut the image according to the ROI, obtain the processed swing image and mark it, build the swing detection model, which is a target detection model. It uses a neural network. The model performs swing detection on the swing image, compares the prediction result with the mark, trains the swing detection model, and finally obtains the trained swing detection model. After Gaussian processing and cutting the image, the image to be detected is generated, which is input into the trained model for swing detection to obtain the detection result, including the position, width and height of the silk thread swing in the image.

[0071] Step 4, judging whether the thread swing is abnormal, specifically, obtaining an average thread swing width according to the number of threads, and then comparing the thread swing width of each thread with the average value to obtain a thread swing exceeding the average value as an abnormal swing.

[0072] like Fig. 9 As shown, point 4: the second camera captures the swing image of the silk thread on the third winding shaft and the fourth winding shaft. The detection method is the same as point 3 above, except that the second camera is 1700mm-1800mm above the ground and is located to the upper right of the third wire guide roller. The second camera is used to capture the swing of the silk thread on the third winding shaft and the fourth winding shaft.

[0073] like Figure 10-12 As shown, point five: the third camera captures the image of the silk thread between the first wire guide roller and the second wire guide roller.

[0074] Step 1: The AGV carrying the collaborative robot moves to the aisle in front of the winding shaft. At this time, the third camera is 1500mm-1700mm from the ground and is located between the first wire guide roller and the second wire guide roller. The third camera is used to capture the wire jump image.

[0075] Step 2: Use the rotation matrix to rotate the collected wire skipping image so that the wire is in a vertical state in the image; the expression of the rotation matrix is ​​as follows:

[0076]

[0077] In the formula, HomMat2D is the generated empty matrix, Phi in the R matrix is ​​the angle to be rotated, and HomMat2DRotate is the final rotation matrix.

[0078] After the image is rotated, a ROI is predefined, and threshold segmentation is performed within the ROI area to separate the second godet fixed end area from the background area.

[0079] Step 3, find the eigenvalue of the row index row2 in the lower right corner of the fixed end area, further locate the detection area of ​​the silk thread based on the eigenvalue of row2, analyze the area, and find the corresponding eigenvalue according to the shooting position. Due to the limited depth of field of the camera, the third camera has four shooting positions because it is not possible to shoot all 80 silk threads at one time. The first shooting position shoots 1-20 silk threads clearly, the second shooting position shoots 15-35 silk threads clearly, the third shooting position shoots 30-65 silk threads clearly, and the fourth shooting position shoots 55-80 silk threads clearly. When in the first shooting position, find the eigenvalue of column1 in the upper left corner of the silk thread area, and when in the second, third, and fourth shooting positions, find the eigenvalue of column2 in the lower right corner of the silk thread area.

[0080] Step 4: Generate the final detection area based on the above feature values.

[0081] Step 5: Obtain the processed wire skipping images and classify and annotate them, build a wire skipping defect detection model, perform defect detection on the spandex yarn images, and compare the predicted defect classification with the actual classification annotations to train the wire skipping defect detection model, and finally obtain a trained wire skipping defect detection model. The spandex yarn images to be detected processed in steps 2-4 are input into the trained wire skipping defect detection model, wire skipping defect detection is performed, and the detection results are obtained.

[0082] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0083] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0084] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0085] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0086] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for detecting thread defects in the production of a spandex textile machine, wherein each spandex textile machine comprises a first winding shaft, a second winding shaft, a third winding shaft, a fourth winding shaft, a first wire guide roller, a second wire guide roller and a third wire guide roller, and upper and lower pressure rollers are arranged between the first winding shaft and the second winding shaft, and between the third winding shaft and the fourth winding shaft, and a plurality of threads conveyed by the first wire guide roller are evenly wound on the first winding shaft, the second winding shaft, the third winding shaft and the fourth winding shaft through the second wire guide roller and the third wire guide roller; characterized in that: The method includes: Four light sources are arranged on each spandex textile machine, and an AGV is used as the inspection carrier of the collaborative robot. Three cameras with different focal lengths are installed on the robotic arm of the collaborative robot, which are the first camera, the second camera, and the third camera in ascending order of focal length. The first camera takes images of two sets of upper and lower pressure rollers, and the system performs reverse winding defect detection on the images; The first camera captures the swing images of the silk thread on the first winding shaft and the second winding shaft, and the second camera captures the swing images of the silk thread on the third winding shaft and the fourth winding shaft, and the system performs swing abnormality detection on the swing images; The third camera captures an image of the wire between the first wire guide roller and the second wire guide roller, and the system performs wire skipping defect detection on the image.

2. The method for detecting thread defects in spandex textile machine production according to claim 1, characterized in that: Four light sources are installed on the side of the textile machine, and each light source is located above a winding shaft, used to illuminate the entire row of silk threads conveyed to the winding shaft. The light source is emitted from the transmitting end to the fixed end face of the winding shaft; the focal length of the first camera is 8mm, the focal length of the second camera is 12mm, and the focal length of the third camera is 25mm.

3. The method for detecting thread defects in spandex textile machine production according to claim 2, characterized in that: The emitting end of the light source is a circular emitting port with a diameter of 55mm-100mm, which is focused by a plano-concave mirror to form a bright band.

4. The method for detecting thread defects in spandex textile machine production according to claim 1, characterized in that: The first camera takes images of each set of upper and lower rollers, and the system performs reverse winding defect detection on the images, including: The AGV with the collaborative robot moves to the aisle in front of the winding shaft. At this time, the first camera is 1700mm-1800mm away from the ground, and the first camera is used to capture the image of the pressure roller. Perform reverse winding defect detection on the dark roller in the image and extract the reverse winding defect using background difference method; In the detection of the roller reverse winding defect with bright image, the background difference method is also first used to obtain the pending reverse winding defect area. Since the brightness of this roller is high and if there is interference from shaft dirt, further processing is required to distinguish true and false defects: first, the roller is extracted from the image using threshold segmentation. When the pending reverse winding defect is surrounded by a black background, it is determined to be a true reverse winding defect. When there is a certain gray value around the pending reverse winding defect, it is determined to be a false reverse winding defect. Finally, the reverse winding defect occurs at the upper pressure roller and / or the lower pressure roller.

5. The method for detecting thread defects in spandex textile machine production according to claim 1, characterized in that: The first camera captures the swing image of the silk thread on the first winding shaft and the second winding shaft, and the system performs swing abnormality detection on the swing image, including: The AGV with the collaborative robot moves to the aisle in front of the winding shaft. At this time, the first camera is 1500mm-1700mm from the ground and is located to the right of the third wire guide roller. The first camera is used to capture the swing image. Perform Gaussian filtering on the acquired swing image; A swing detection model is constructed using a neural network, and the model is trained. The image to be detected after Gaussian processing is input into the trained swing detection model for swing detection to obtain the detection result, i.e., the position, width and height of the silk thread swing in the image; To judge whether the thread swing is abnormal, specifically, the average thread swing width is obtained according to the number of threads, and then the thread swing width of each thread is compared with the average value to obtain the thread swing exceeding the average value as abnormal swing.

6. The method for detecting thread defects in spandex spinning machine production according to claim 5, characterized in that: The second camera captures the swing images of the silk thread on the third winding shaft and the fourth winding shaft. When the system performs swing anomaly detection on the swing images, the second camera is 1700mm-1800mm from the ground and is located to the upper right of the third wire guide roller.

7. The method for detecting thread defects in spandex textile machine production according to claim 1, characterized in that: The third camera captures the image of the wire between the first wire guide roller and the second wire guide roller, and the system performs wire skipping defect detection on the image, including: The AGV with the collaborative robot moves to the aisle in front of the winding shaft. At this time, the third camera is 1500mm-1700mm from the ground and is located between the first wire guide roller and the second wire guide roller. The third camera is used to capture the wire jump image. The collected skipped wire image is rotated using a rotation matrix so that the wire is in a vertical state in the image; After rotating the image, threshold segmentation is performed in the ROI area to separate the second godet fixed end area from the background; Calculate the eigenvalue of the row index row2 at the lower right corner of the fixed end area, further locate the detection area of ​​the silk thread based on the eigenvalue of row2, analyze the area again, and calculate the corresponding eigenvalue according to the shooting position; Generate the final detection area based on the above eigenvalues; A wire skipping defect detection model is constructed using a neural network, and the model is trained. The processed images to be detected are input into the trained wire skipping defect detection model for wire skipping defect detection to obtain the detection results.

Images