Textile material laser radar image enhancement type broken line identification method and system

Abstract

The invention discloses a textile material laser radar image enhancement type broken line identification method and system, particularly relates to the technical field of laser radars, and aims to solve the problems of limited monitoring range and insufficient detection efficiency in the existing broken line detection technology. According to the method, laser pulse scanning is conducted on a spinning area from the lower end of a front roller of a spinning frame to a steel collar through a laser radar, yarn movement change information is collected in real time, a scanning mode of laser pulses is recorded, and in an image processing server, a laser feature change rate and a target feature change rate are generated according to the scanning mode and the yarn movement information; according to the method, matching verification is carried out to evaluate matching, authenticity of a laser radar image is checked, denoising enhancement and standardization processing are carried out on the image, texture features and reflection intensity fluctuation features are extracted, and broken yarn identification is carried out by using a deep learning method, so that efficient and accurate detection of spun yarn broken yarns of a textile material can be realized. The broken line identification efficiency is improved; and the computing power resource consumption is reduced.

Description

Technical Field

[0001] The present invention relates to the field of laser radar technology, and more specifically, to a method and system for identifying broken lines in textile materials using laser radar images. Background Art

[0002] During the yarn drafting process, a broken thread failure may occur, resulting in interruption of the drafting and affecting the production of spun yarn. The existing broken thread identification method based on visual detection sets a mobile optical camera to shoot and image the yarn area, and identifies and processes the yarn image to determine the broken thread failure. However, it is difficult for the optical camera to monitor and process multiple yarn areas in real time, and the mobile start-stop control method has a complex structure and is prone to failure. The failure stagnation time of the broken thread detection is too long, which affects the recovery timeliness of the spun yarn production. LiDAR technology is used to detect yarn breakage. If the material properties of multiple yarns are different, the reflection efficiency of different yarns to lasers will be different. The photoelectric detector used to capture the reflected laser will find it difficult to identify different yarns and will not be able to effectively identify and judge the yarn breakage fault. If the difference in the running speed of the yarn in the spinning machine is superimposed, the laser scanning will produce dynamic blur, which increases the difficulty of comparing and judging the LiDAR imaging, making it difficult to effectively identify the yarn breakage. At the same time, the continuous processing of the real-time transmitted image data puts a heavy pressure on the computing resources of the breakage recognition system.

[0003] In order to solve the above defects, a technical solution is now proposed. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method and system for identifying broken wires of textile materials using laser radar image enhancement, and the specific steps include: Step S1, performing laser pulse scanning on the target area through the laser radar, collecting the yarn movement change information in the target area, and recording the scanning mode of the laser pulse, and uploading the laser radar image data to the image processing server in real time in combination with the network; the target area is the spinning area from the lower end of the front roller of the spinning frame to the steel ring; Step S2, in the image processing server, a laser feature change rate is generated according to the scanning mode of the laser pulse, a target feature change rate is generated according to the yarn motion change information, the laser feature change rate and the target feature change rate are verified for consistency, and the matching of the laser pulse scanning and the yarn motion is evaluated; Step S3, verifying the authenticity of the laser radar image according to the matching evaluation result of the laser pulse scanning and the yarn movement, and establishing a confidence model according to the authenticity test result of the known laser radar image, performing a confidence test on the laser radar image acquired in real time by the image processing server, and adding a confidence label to the laser radar image according to the result of the confidence test, wherein the confidence label includes a high confidence image and a low confidence image; Step S4, denoising, enhancing and standardizing the high confidence image to generate an image to be processed, extracting texture features and reflection intensity fluctuation features in the image to be processed by combining the edge detection method, fusing the texture features and reflection intensity fluctuation features and converting them into feature vectors, using a deep learning method to identify the broken line image, and verifying the broken line identification by combining the confidence label and authenticity label of the image to be processed.

[0005] In a preferred embodiment, step S1 includes the following contents: S1-1, determine the target area of ​​the laser radar, including the location, distribution and quantity of the spun yarn in the target area , the target area is divided into several sub-areas, each of which corresponds to the laser scanning of a spun yarn; S1-2, determine the parameters of the laser radar and the characteristic information of the yarn, including the wavelength range of the laser radar , power range , speed range , the reflectivity of each yarn , the speed of each spun yarn ,in is the index of the spun yarn; S1-3, during the laser radar scanning process, the real-time wavelength and power of the laser radar are recorded as the scanning mode, the reflectivity and movement speed of the yarn in the target area are recorded through laser scanning as the yarn movement change information, the scanning mode and the yarn movement information are timestamped, and the real-time acquired laser radar image and the scanning mode and yarn movement information with the corresponding timestamp are uploaded to the image processing server through the configured network.

[0006] In a preferred embodiment, step S2 includes the following contents: S2-1, according to the scanning mode of the laser pulse, the wavelength of the marking laser pulse when irradiating the previous sub-area is , the laser pulses are separated by a time interval The wavelength when the next sub-area is illuminated is , then the laser pulses are separated by a time interval The wavelength change rate after ; S2-2, according to the scanning mode of the laser pulse, the power of the marking laser pulse when irradiating the previous sub-area is , the laser pulses are separated by a time interval The power when irradiating the next sub-area is , then the laser pulses are separated by a time interval The power change rate after ; S2-3, by defining the influence function of the power and wavelength of the laser scanning on the yarn reflection intensity, the influence degree of the change of the laser adjustment logic on the target area is evaluated to obtain the laser scanning influence coefficient; S2-4, the laser characteristic change rate is obtained according to the wavelength change rate, power change rate and laser scanning influence coefficient of the laser pulse. The calculation expression of the laser characteristic change rate is: ; S2-5, according to the yarn movement change information, the reflectivity and movement speed of each yarn are converted into a time domain signal, and the reflectivity of the yarn at the time is calibrated. for The yarn movement speed is for , and and Perform filtering, denoising and standardization processing; S2-6, use the wavelet decomposition method to decompose the time domain signal after filtering, denoising and standardization, perform wavelet decomposition on the reflectivity of the yarn to obtain the high-frequency coefficients of the reflectivity signal, and calculate the reflectivity change rate , where and Separately for the moment and time The high-frequency coefficient of the reflectivity signal is obtained by wavelet decomposition of the yarn movement speed, and the change rate of the movement speed is calculated as , where and Separately for the moment and time The high frequency coefficient of the motion speed signal; S2-7, obtain the target feature change rate according to the reflectivity change rate and the motion speed change rate. The expression of the target feature change rate is: , where is the target feature change rate, are the weight coefficients of the reflectivity change rate and the motion speed change rate, respectively, and All are positive numbers; S2-8, by calculating the laser characteristic change rate and the target feature change rate The Pearson correlation coefficient between the two is used for correlation verification. The calculation expression of the Pearson correlation coefficient is: , where is the laser characteristic change rate and the target feature change rate The Pearson correlation coefficient between is the laser characteristic change rate and the target feature change rate The covariance between and are the laser characteristic change rates and the target feature change rate The standard deviation of the laser characteristic change rate is calculated by and the target feature change rate Euclidean distance to evaluate the laser feature change rate and the target feature change rate The difference between , the calculation expression of Euclidean distance is ; S2-9, laser feature change rate according to Pearson correlation coefficient and Euclidean distance and the target feature change rate The fit is evaluated, and the fit expression is , set the fit threshold to If the fit Greater than or equal to the fit threshold , then the laser scan of the mark matches the yarn motion characteristics. If the match is Less than the fit threshold , then the marker laser scan deviates from the yarn motion characteristics.

[0007] In a preferred embodiment, in S2-3, the influence function of the power and wavelength of the laser scanning on the yarn reflection intensity is defined to evaluate the influence of the laser adjustment logic change on the target area, and the logic for obtaining the laser scanning influence coefficient is: S2-3-1, calibrate the wavelength influence function to characterize the influence of the laser wavelength on the target area when the laser adjustment logic changes. Use the quadratic polynomial model to define the wavelength influence function, then the wavelength influence function is , where is the yarn reflection intensity obtained by laser scanning, is the wavelength of the laser scan, where As the model parameters, the quadratic polynomial model is fitted using the least squares method, and the error function is defined as , where For the The actual reflection intensity of the yarn in the scan, is the number of laser scans, is the yarn reflection intensity predicted by the model. about Find the partial derivative and set it to 0, and get the wavelength equation as , solve the wavelength equation to obtain the model parameters , and substitute it into the quadratic polynomial model to obtain the wavelength influence function ; S2-3-2, calibrate the power influence function to characterize the influence of laser power on the target area when the laser adjustment logic changes. Use a logarithmic model to define the power influence function, then the power influence function is , where is the yarn reflection intensity obtained by laser scanning, is the power of laser scanning, where As the model parameters, the least squares method is used to fit the logarithmic model, and the error function is defined as , where For the The actual reflection intensity of the yarn in the scan, is the yarn reflection intensity predicted by the model. about Find the partial derivative and set it to 0, and the power equation is obtained as , solve the power equation to obtain the model parameters , and substitute it into the logarithmic model to obtain the power influence function: ; S2-3-3, the laser scanning influence coefficient is generated based on the wavelength influence function and the power influence function. The expression of the laser scanning influence coefficient is: .

[0008] In a preferred embodiment, step S3 includes the following contents: S3-1, verifying the authenticity of the laser radar image according to the consistency verification. If the laser scan matches the yarn motion feature, a true image label is added to the laser radar image transmitted in real time at the corresponding timestamp. If the laser scan deviates from the yarn motion feature, a false image label is added to the laser radar image transmitted in real time at the corresponding timestamp. S3-2, define the lidar image with the real image label added as 0, and the lidar image with the fake image label added as 1, collect the true and false labels of the lidar image transmitted in real time within the time period T and generate the true and false sequence. The true and false sequence is formed by the combination of 0 and 1 arranged in time order. The number of 0 in the true and false sequence is calibrated as , the total number of 0s and 1s in the true and false sequences is , then the quality coefficient of the true and false sequence is calculated as , the preset quality threshold is , if the quality coefficient Less than or equal to the quality threshold , the output device of the image processing server will prompt the staff to adjust the logic misalignment of the laser scanning and generate confidence adjustment information. If the quality coefficient Greater than quality threshold , the output device of the image processing server prompts the staff to adjust the logic synchronization of the laser scanning, and no confidence adjustment information is generated.

[0009] In a preferred embodiment, the method for verifying the authenticity of the laser radar image in step S3 is: S3-1-1, based on machine learning methods, a confidence model is established for authenticity verification of LiDAR images. The binary classification model is trained using known historical LiDAR image data. The binary classification model is established using the logistic regression method, and the expression is: , where the degree of fit is the input feature, is the probability that the lidar image is a true image, is the weight of the binary classification model, is the bias term of the binary classification model. For a known historical lidar image data, the loss function is defined as , for all known historical lidar image data, calculate the average value of the loss function and use the gradient descent method to optimize the model parameters until the loss function converges or the binary classification model reaches the maximum number of iterations; S3-1-2, use the trained binary classification model to detect the real-time transmitted LiDAR image. If the value of is greater than or equal to the preset authenticity verification threshold, the true image label is added to the lidar image. If the value of is less than the preset authenticity verification threshold, a false image label is added to the lidar image.

[0010] In a preferred embodiment, the logic for generating the confidence adjustment information in step S3 is: S3-2-1, collect the yarn reflection intensity signal in the target area , according to the yarn reflectivity of the target area , adjust the laser pulse wavelength to ; S3-2-2, according to the yarn reflection signal intensity in the target area, adjust the laser pulse power to , where is the power adjustment factor, is the adjusted target signal strength; S3-2-3, according to the movement speed of the spinning , adjust the laser radar scanning speed to , where is the maximum scanning speed of the laser radar, is the maximum movement speed of the yarn; S3-2-4, generating confidence adjustment information according to adjusting the laser pulse wavelength, adjusting the laser pulse power and adjusting the laser radar scanning speed: .

[0011] In a preferred embodiment, step S4 includes the following: S4-1, using filtering to remove background noise in the high confidence image, using histogram equalization to enhance the contrast in the high confidence image, and standardizing the grayscale value of the high confidence image; S4-2, using an edge detection operator to calculate the gradient of the preprocessed high confidence image, detecting the contour features of the yarn in the target area, and obtaining the reflection intensity features according to the average reflection intensity of the yarn contour position; S4-3, establish a convolutional neural network model and use the extracted contour features and reflection intensity features for training, generate a labeled data set based on the known high-confidence image data, each data sample in the labeled data set includes input features and broken line labels, the input features include contour features and reflection intensity features, and the broken line labels are divided into "broken line" and "not broken line" as the output of the convolutional neural network model; S4-4, concatenate the contour feature and the reflection intensity feature into a vector As the input of the convolutional neural network model, the vector The local features are extracted by the convolution layer of the convolutional neural network model, flattened and output into a one-dimensional vector, which enters the fully connected layer for classification, and the disconnection label is output through the output layer. The activation function of the fully connected layer is ReLU, and the activation function of the output layer is sigmoid. S4-5, define the loss function as ,in, is the predicted output of the convolutional neural network model, The convolutional neural network model parameters are optimized by minimizing the loss function, and the model probability value is output through the sigmoid activation function. , where is the output of the last layer of the convolutional neural network; S4-6, define the output threshold as , the model probability value With output threshold For comparison, if the model probability value Greater than or equal to the output threshold , then the image processing server outputs the recognition result through the network as disconnected. If the model probability value Less than the output threshold , then the image processing server outputs the recognition result through the network as being not disconnected; S4-7, after performing disconnection recognition on the high-confidence image transmitted in real time, using the idle computing resources of the image processing server to extract the lidar image with the true image label and the low-confidence image label, and inputting the lidar image data with the true image label and the low-confidence image label into the convolutional neural network model for disconnection recognition, and saving the disconnection recognition results in timestamp order, and deleting all lidar image data with false image labels; S4-8, if a yarn break is identified in the high-confidence image transmitted in real time or the laser radar image with a true image label and a low-confidence image label, the image processing server output device prompts the staff that the yarn is broken.

[0012] In a preferred embodiment, the method for acquiring the profile characteristics and reflection intensity characteristics of the yarn in step S4 is: S4-2-1, use edge detection operator to calculate the horizontal gradient of the image and the gradient in the vertical direction , the gradient amplitude is calculated as and gradient direction , perform non-maximum suppression on each pixel on the image, and retain strong and weak edges by setting high and low thresholds to obtain edge features of high-confidence images; S4-2-2, define the yarn edge area using a sliding window based on the obtained edge features , calculate the yarn edge area The average reflection intensity of all pixels in the image is calculated as , where is the reflection intensity of all pixels in the yarn edge area, Yarn edge area The total number of pixels.

[0013] A laser radar image enhanced broken line recognition system for textile materials, including a laser scanning module, a server filtering module, a laser radar image processing module, a feature extraction module, and a deep learning module; The laser scanning module is used to perform laser pulse scanning on the target area, collect the movement change information of the yarn in the target area, and record the scanning mode of the laser pulse, and upload the laser radar image data to the image processing server in real time in combination with the network; the target area is the spinning area from the lower end of the front roller of the spinning frame to the steel ring; The server filtering module is used to receive and process the laser radar image data transmitted by the laser scanning module, specifically including: generating a laser characteristic rate of change according to a scanning pattern of the laser pulse; Generate target feature change rate according to the motion change information of the yarn; Verify the consistency between the laser feature change rate and the target feature change rate, and evaluate the matching between the laser pulse scanning and the yarn movement; The authenticity of the laser radar image is checked according to the matching evaluation result, a confidence model is established, a confidence prior is performed on the laser radar image acquired in real time, and a confidence label is added to the image, wherein the confidence label includes a high confidence image and a low confidence image; The laser radar image processing module is used to perform denoising, enhancement and standardization processing on high-confidence images to generate images to be processed, and extract the reflection intensity fluctuation characteristics in the images to be processed in combination with edge detection methods; A feature extraction module is used to fuse the reflection intensity fluctuation feature with the edge feature and convert it into a feature vector; The deep learning module is used to identify the broken line image using the deep learning method based on the feature vector, and verify the broken line identification result according to the confidence label and authenticity label of the image to be processed.

[0014] In the above technical solution, the technical effects and advantages provided by the present invention are: 1. The present invention uses a laser radar to perform laser pulse scanning on the spun yarn area of ​​the textile material, collects the movement change information of the yarn in real time, and combines the scanning mode of the laser pulse for accurate analysis. By comparing the consistency between the laser feature change rate and the target feature change rate, the time point when the wire break occurs is accurately identified, effectively avoiding the problem of long detection stagnation time caused by the complex start-stop control of the camera movement in the traditional method; 2. The present invention dynamically adjusts the scanning parameters of the laser radar, which can adapt to the changes in the reflection characteristics and movement speed of the yarn, ensure the consistency and high quality of image acquisition, use wavelet transform technology to analyze the reflectivity of the yarn and the rate of change of the movement speed, improve the detection sensitivity of the laser radar to the broken yarn, identify the broken thread image through a deep learning method, and verify the identification result in combination with the confidence label and the authenticity label, further enhancing the robustness and accuracy of the detection. In the process of broken thread identification, the idle computing resources of the image processing server are used to further analyze the laser radar images with true image labels and low confidence image labels, thereby improving the utilization efficiency of resources. By establishing a confidence model and dynamically adjusting the scanning parameters, the reflection signal can be effectively captured under different yarn movement states, thereby improving the efficiency of broken thread detection. The present invention is not only suitable for various types of yarns and different reflection characteristics, but also can cope with the changes in yarn movement speed and the dynamic blur problem of acquired images, and has strong adaptability and a wide range of applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0016] Figure 1 The figure is a flow chart of the method of the present invention.

[0017] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Example 1: Please refer to Figure 1 As shown, the present invention is a method for identifying broken lines of textile materials using laser radar image enhancement, and the specific steps include: Step S1, performing laser pulse scanning on the target area through the laser radar, collecting the yarn movement change information in the target area, and recording the scanning mode of the laser pulse, and uploading the laser radar image data to the image processing server in real time in combination with the network; the target area is the spinning area from the lower end of the front roller of the spinning frame to the steel ring; Step S2, in the image processing server, a laser feature change rate is generated according to the scanning mode of the laser pulse, a target feature change rate is generated according to the yarn motion change information, the laser feature change rate and the target feature change rate are verified for consistency, and the matching of the laser pulse scanning and the yarn motion is evaluated; Step S3, verifying the authenticity of the laser radar image according to the matching evaluation result of the laser pulse scanning and the yarn movement, and establishing a confidence model according to the authenticity test result of the known laser radar image, performing a confidence test on the laser radar image acquired in real time by the image processing server, and adding a confidence label to the laser radar image according to the result of the confidence test, wherein the confidence label includes a high confidence image and a low confidence image; Step S4, denoising, enhancing and standardizing the high confidence image to generate an image to be processed, extracting texture features and reflection intensity fluctuation features in the image to be processed by combining the edge detection method, fusing the texture features and reflection intensity fluctuation features and converting them into feature vectors, using a deep learning method to identify the broken line image, and verifying the broken line identification by combining the confidence label and authenticity label of the image to be processed.

[0020] Step S1 includes the following contents: S1-1, determine the target area of ​​the laser radar, including the location, distribution and quantity of the spun yarn in the target area , the target area is divided into several sub-areas, each of which corresponds to the laser scanning of a spun yarn; S1-2, determine the parameters of the laser radar and the characteristic information of the yarn, including the wavelength range of the laser radar , power range , speed range , the reflectivity of each yarn , the speed of each spun yarn ,in is the index of the spun yarn; The reflectivity and movement speed of each yarn are obtained according to the results of pre-calibration or real-time detection. For example, the reflectivity of each yarn is obtained by the pre-calibration method: using a known wavelength and known power The laser pulse scans the standard material plate in an optical darkroom. is the index of the number of laser scans, and the reflectivity of the standard material plate is , get the echo signal strength of the standard plate , place the yarn in the target area using the same known wavelength and known power Scan the yarn in an optical darkroom to obtain the echo signal intensity of the yarn , then the reflectivity of the spun yarn is , the reflectivity of each spun yarn is obtained and the corresponding wavelength Keep records; The method of obtaining the speed of the spinning yarn by real-time detection is: using a laser radar to detect the position of the spinning yarn in the target area through the echo signal of the laser pulse , record the time interval of continuous scanning And fine yarn in Position changes within , calculate the spinning speed as , use the sliding average filter to smooth the speed data to obtain the smoothed spinning speed , where is the speed of the smoothed yarn, is the number of sliding windows, is the index of the sliding window.

[0021] S1-3, during the laser radar scanning process, the real-time wavelength and power of the laser radar are recorded as the scanning mode, the reflectivity and movement speed of the yarn in the target area are recorded through laser scanning as the yarn movement change information, the scanning mode and the yarn movement information are timestamped, and the real-time acquired laser radar image and the scanning mode and yarn movement information with the corresponding timestamp are uploaded to the image processing server through the configured network.

[0022] The wavelength and power of the laser radar are adjusted in real time when scanning different yarns in the target area to obtain a laser radar image for identifying and distinguishing each yarn. The specific adjustment scanning logic of the laser radar wavelength and power depends on the professional and technical personnel in this field to set it according to the yarn reflectivity characteristics, movement speed and background noise level and other factors, which will not be elaborated here; By marking the timestamps for the laser radar scanning mode and the yarn movement change information, the data collected by the laser radar image is synchronized with the actual movement state of the yarn. The timestamp marking can accurately track the relationship between the image data obtained by the laser radar scanning and the position, speed, reflection intensity and other characteristics of the yarn at a specific time point. The movement of the yarn has high-speed and complex dynamic changes. Through the timestamp marking, all data are ensured to have a unified timing standard, thereby avoiding data confusion; Depending on the different characteristics of the yarn, such as reflectivity, material, movement speed, etc., the parameters of the LiDAR may need to be adjusted in real time. Parameters such as wavelength, power, scanning speed, etc. If the changes in the yarn are not monitored and recorded in real time, the scanning parameters of the LiDAR may not be able to adapt to the changes in the yarn state, resulting in unstable image quality or failure of broken wire detection.

[0023] Step S2 includes the following contents: S2-1, according to the scanning mode of the laser pulse, the wavelength of the marking laser pulse when irradiating the previous sub-area is , the laser pulses are separated by a time interval The wavelength when the next sub-area is illuminated is , then the laser pulses are separated by a time interval The wavelength change rate after ; S2-2, according to the scanning mode of the laser pulse, the power of the marking laser pulse when irradiating the previous sub-area is , the laser pulses are separated by a time interval The power when irradiating the next sub-area is , then the laser pulses are separated by a time interval The power change rate after ; S2-3, by defining the influence function of the power and wavelength of the laser scanning on the yarn reflection intensity, the influence degree of the change of the laser adjustment logic on the target area is evaluated to obtain the laser scanning influence coefficient; The method for obtaining the laser scanning influence coefficient is: S2-3-1, calibrate the wavelength influence function to characterize the influence of the laser wavelength on the target area when the laser adjustment logic changes. Use the quadratic polynomial model to define the wavelength influence function, then the wavelength influence function is , where is the yarn reflection intensity obtained by laser scanning, is the wavelength of the laser scan, where As the model parameters, the quadratic polynomial model is fitted using the least squares method, and the error function is defined as , where For the The actual reflection intensity of the yarn in the scan, is the number of laser scans, is the yarn reflection intensity predicted by the model. about Find the partial derivative and set it to 0, and get the wavelength equation as , solve the wavelength equation to obtain the model parameters , and substitute it into the quadratic polynomial model to obtain the wavelength influence function ; S2-3-2, calibrate the power influence function to characterize the influence of laser power on the target area when the laser adjustment logic changes. Use a logarithmic model to define the power influence function, then the power influence function is , where is the yarn reflection intensity obtained by laser scanning, is the power of laser scanning, where As the model parameters, the least squares method is used to fit the logarithmic model, and the error function is defined as , where For the The actual reflection intensity of the yarn in the scan, is the yarn reflection intensity predicted by the model. about Find the partial derivative and set it to 0, and the power equation is obtained as , solve the power equation to obtain the model parameters , and substitute it into the logarithmic model to obtain the power influence function: ; S2-3-3, the laser scanning influence coefficient is generated based on the wavelength influence function and the power influence function. The expression of the laser scanning influence coefficient is: ; Theoretically, the influence of laser radar scanning on the yarn area can be measured by calculating the coverage and scanning efficiency of laser scanning. In practical applications, coverage and scanning efficiency can directly reflect the effect of laser scanning. However, the reason why the "influence function of laser power and wavelength on yarn reflection intensity" is introduced is that this method is more refined and can more directly quantify the influence of laser scanning process on yarn reflection. Coverage refers to the size of the area irradiated by the laser beam in the laser scanning area. For yarn, the area size will be affected by many factors, including the shape of the yarn, the distribution of the yarn in the scanning area, the surface reflectivity of the yarn, etc. The change of the surface characteristics of the yarn will affect the intensity of the reflected light. Therefore, even if the laser coverage is large, the intensity distribution of the light may not be uniform, resulting in different scanning effects; The surface characteristics and conditions of the yarn, such as wetness, roughness, coating, etc., will affect the ability to reflect light. If the impact is simply measured by coverage, it will not be able to fully capture the details of the surface characteristics and conditions of the yarn; Scanning efficiency refers to the speed and quality of effective information collection during the laser radar scanning process. It is affected by factors such as scanning speed, yarn movement speed, and reflectivity distribution. If the impact of laser is evaluated only by scanning efficiency, it is difficult to intuitively understand the contribution of light intensity to broken wire detection. For different types of yarns, scanning efficiency cannot reflect the true impact of laser pulses on broken wire detection. Therefore, it is difficult to directly calculate the coverage and scanning efficiency to accurately reflect the complex effects of laser wavelength and power changes on yarn reflection; The reflectivity of yarn is different for lasers of different wavelengths. For example, fiber materials have higher reflectivity for lasers of certain wavelengths, but weaker reflectivity for lasers of other wavelengths. By calculating the effect of wavelength and power on reflection intensity, we can more accurately understand the changes in LiDAR signals. The effect of wavelength and power on reflection intensity directly reflects the accuracy of broken wire detection. This approach takes into account the complex relationship between laser wavelength, power and yarn reflectance properties, and therefore enables a more accurate assessment of the actual impact of laser scanning on the yarn; The yarn break detection depends on the change of reflection intensity. When the yarn breaks, the reflection intensity will change significantly. Therefore, the reflection intensity is one of the core indicators of the yarn break detection. By accurately calculating the influence of laser wavelength and power changes on the reflection intensity, the detection capability of laser scanning under different conditions can be more effectively evaluated. The change of reflection intensity is not only related to the wavelength and power of the laser, but also closely related to the reflection characteristics of the yarn. The change of reflection intensity provides more detailed levels for evaluating the accuracy of broken thread detection. Laser wavelength, power and yarn reflectivity change dynamically. The reflectivity of yarn changes over time. LiDAR needs to continuously adjust the wavelength and power according to the reflection characteristics of the yarn. Direct calculation of coverage and scanning efficiency cannot fully capture this dynamic change. By using the influence function of reflection intensity, the actual impact of the laser on the yarn can be fed back in real time, and the scanning parameters can be dynamically adjusted to ensure the accuracy of broken thread detection.

[0024] Calculating the coverage and scanning efficiency of the laser on the yarn area is an important measurement indicator, but the coverage and scanning efficiency consider the range of the laser scanning area and the scanning speed from a geometric perspective. In order to achieve more accurate and reliable broken wire detection, the interaction between the laser and the yarn is considered from the perspective of physics and signal processing, that is, the reflection intensity is affected by adjusting the wavelength and power, thereby affecting the final detection effect. Therefore, the introduction of the influence function of the laser wavelength and power on the yarn reflection intensity can provide a quantitative tool that more accurately reflects the influence of laser scanning on yarn broken wire detection; Although the impact of laser scanning on the yarn area can be quantified by directly calculating the coverage and scanning efficiency, it cannot accurately reflect the specific impact of laser wavelength and power changes on the yarn reflection intensity. Therefore, the method of introducing the "influence function of laser wavelength and power on yarn reflection intensity" to obtain the laser scanning influence coefficient is adopted to more carefully characterize the impact of laser scanning on broken wire detection, provide more accurate measurement and optimization basis, and comprehensively consider the complex physical relationship between laser and yarn, thereby improving the effectiveness and stability of broken wire detection.

[0025] S2-4, the laser characteristic change rate is obtained according to the wavelength change rate, power change rate and laser scanning influence coefficient of the laser pulse. The calculation expression of the laser characteristic change rate is: ; S2-5, according to the yarn movement change information, the reflectivity and movement speed of each yarn are converted into a time domain signal, and the reflectivity of the yarn at the time is calibrated. for The yarn movement speed is for , and and Perform filtering, denoising and standardization processing; S2-6, use the wavelet decomposition method to decompose the time domain signal after filtering, denoising and standardization, perform wavelet decomposition on the reflectivity of the yarn to obtain the high-frequency coefficients of the reflectivity signal, and calculate the reflectivity change rate , where and Separately for the moment and time The high-frequency coefficient of the reflectivity signal is obtained by wavelet decomposition of the yarn movement speed, and the change rate of the movement speed is calculated as , where and Separately for the moment and time The high frequency coefficient of the motion speed signal; By converting the reflectivity and movement speed of each yarn into time domain signals for wavelet decomposition, the dynamic changes of the yarn reflectivity and movement speed can be observed simultaneously at multiple scales, which can not only capture large slow changes but also identify small rapid fluctuations. The advantage of wavelet transform is its multi-scale analysis capability, which can capture the local change characteristics of the signal. By analyzing the high-frequency parts at different scales, the change rates of reflectivity and movement speed at different scales can be obtained. For example, low-frequency changes may represent slower trend changes, while high-frequency changes may represent instantaneous fluctuations. By comprehensively considering the changes at different scales, the disconnection anomaly can be identified more accurately. The change rate of reflectivity and movement speed obtained by wavelet transform is used to analyze the degree of fit between the laser scanning effect and the yarn change. Specifically, if the laser radar can accurately capture the change of the yarn, that is, the change of reflectivity and movement speed, the fit is verified by comparing the difference between the reflectivity and the data collected by the laser radar; For example, the laser radar should be able to capture significant fluctuations in the signal when the yarn reflectivity changes greatly, or quickly adjust scanning parameters such as wavelength and power when the yarn moves quickly. By comparing the multi-scale analysis of the yarn feature changes and the laser radar response, the laser scanning parameters can be adjusted to adapt to the movement and reflection characteristics of the yarn. Wavelet transform processing improves the laser radar system's sensitivity to yarn breakage detection. The laser radar responds effectively and promptly to slight changes in reflectivity and fluctuations in movement speed, thereby achieving high-precision breakage identification. Wavelet transform effectively removes noise from the data, highlights the core change characteristics of the signal, and improves the robustness and accuracy of laser radar image processing, making it suitable for situations where yarns move at high speed and reflectivity changes greatly. By analyzing the yarn change rate through wavelet transform, the laser radar can adjust the scanning parameters in real time to ensure that the reflected signal is effectively captured under different yarn movement states, thereby improving the efficiency and accuracy of breakage detection.

[0026] S2-7, obtain the target feature change rate according to the reflectivity change rate and the motion speed change rate. The expression of the target feature change rate is: , where is the target feature change rate, are the weight coefficients of the reflectivity change rate and the motion speed change rate, respectively, and All are positive numbers; S2-8, by calculating the laser characteristic change rate and the target feature change rate The Pearson correlation coefficient between the two is used for correlation verification. The calculation expression of the Pearson correlation coefficient is: , where is the laser characteristic change rate and the target feature change rate The Pearson correlation coefficient between is the laser characteristic change rate and the target feature change rate The covariance between and are the laser characteristic change rates and the target feature change rate The standard deviation of the laser characteristic change rate is calculated by and the target feature change rate Euclidean distance to evaluate the laser feature change rate and the target feature change rate The difference between , the calculation expression of Euclidean distance is ; S2-9, laser feature change rate according to Pearson correlation coefficient and Euclidean distance and the target feature change rate The fit is evaluated, and the fit expression is , set the fit threshold to If the fit Greater than or equal to the fit threshold , then the laser scan of the mark matches the yarn motion characteristics. If the match is Less than the fit threshold , then the marking laser scanning deviates from the yarn motion characteristics; By verifying the degree of fit between laser scanning and yarn feature changes, high-quality image data that matches yarn changes can be effectively screened out, irrelevant or noisy images can be removed, and image data that is significantly correlated with yarn feature changes, including reflectivity and movement speed, can be retained to ensure that the image data for subsequent analysis and processing has a higher signal-to-noise ratio and accuracy, reduce redundant data and noise interference, and by screening images that match actual yarn changes, avoid using abnormal data caused by unstable laser scanning parameters, background interference or other factors as valid information, thereby reducing the probability of false detection and missed detection; After screening out the image data related to yarn changes, subsequent analysis only needs to process a small amount of high-quality data, which helps to improve computing efficiency and reduce the burden on the computing system in real-time detection and high-frequency data collection scenarios. The screening process removes unnecessary features and irrelevant data, simplifies subsequent feature extraction, classification and analysis tasks, and avoids feature redundancy caused by excessive data. The reflectivity and movement speed changes of the yarn are directly related to the moment when the thread breaks. By verifying the fit between the laser scan and the changes in yarn characteristics, the time point when the thread breaks can be accurately identified, thereby improving the timeliness and accuracy of thread break detection. The screened image data has high quality and consistency, and can more accurately extract the reflectivity, morphological changes and other features of the yarn, which are crucial for subsequent thread break identification, quality inspection and defect classification tasks.

[0027] Step S3 includes the following contents: S3-1, verifying the authenticity of the laser radar image according to the consistency verification. If the laser scan matches the yarn motion feature, a true image label is added to the laser radar image transmitted in real time at the corresponding timestamp. If the laser scan deviates from the yarn motion feature, a false image label is added to the laser radar image transmitted in real time at the corresponding timestamp. The method for authenticity verification of LiDAR images is as follows: S3-1-1, based on machine learning methods, a confidence model is established for authenticity verification of LiDAR images. The binary classification model is trained using known historical LiDAR image data. The binary classification model is established using the logistic regression method, and the expression is: , where the degree of fit is the input feature, is the probability that the lidar image is a true image, is the weight of the binary classification model, is the bias term of the binary classification model. For a known historical lidar image data, the loss function is defined as , for all known historical lidar image data, calculate the average value of the loss function and use the gradient descent method to optimize the model parameters until the loss function converges or the binary classification model reaches the maximum number of iterations; S3-1-2, use the trained binary classification model to detect the real-time transmitted LiDAR image. If the value of is greater than or equal to the preset authenticity verification threshold, the true image label is added to the lidar image. If the value of is less than the preset authenticity verification threshold, a false image label is added to the lidar image.

[0028] The traditional authenticity verification method based on sliding window requires the accumulation of a certain number of data points. The data will be analyzed only after the window is filled, and the authenticity is judged according to the data in the window. Before the window is filled, the judgment of any new data will be delayed and cannot be processed immediately. Therefore, when the first "true" label lidar image data arrives, the opportunity for timely processing may have been missed. The sliding window method usually has a fixed window size and needs to be analyzed within a fixed window. The window size is pre-set. If the window is large, it may cause data processing delays; if the window is small, there may not be enough statistical information to accurately judge the authenticity of the data. Therefore, the sliding window is difficult to adjust flexibly when processing dynamic data, and it is difficult to handle the data flow of the lidar image with large fluctuations. The sliding window method relies on the historical data in the window for evaluation. In addition, in some cases, the sliding window is affected by the data of the previous period of time, resulting in a lag in the classification results. When the data mutates, the historical data in the window cannot reflect the new data changes in time. In the confidence model-based method, each data point is immediately classified and processed. The confidence model directly calculates the probability that the data point belongs to the "true" category, that is, the confidence, and makes a judgment based on a pre-set threshold. There is no need to wait for the window to fill before analyzing. It can effectively make a judgment as soon as the data arrives and process the lidar image.

[0029] S3-2, define the lidar image with the real image label added as 0, and the lidar image with the fake image label added as 1, collect the true and false labels of the lidar image transmitted in real time within the time period T and generate the true and false sequence. The true and false sequence is formed by the combination of 0 and 1 arranged in time order. The number of 0 in the true and false sequence is calibrated as , the total number of 0s and 1s in the true and false sequences is , then the quality coefficient of the true and false sequence is calculated as , the preset quality threshold is , if the quality coefficient Less than or equal to the quality threshold , the output device of the image processing server will prompt the staff to adjust the logic misalignment of the laser scanning and generate confidence adjustment information. If the quality coefficient Greater than quality threshold , then the output device of the image processing server prompts the staff to adjust the logic synchronization of the laser scanning, and no confidence adjustment information is generated; The logic for generating confidence adjustment information is: S3-2-1, collect the yarn reflection intensity signal in the target area , according to the yarn reflectivity of the target area , adjust the laser pulse wavelength to ; S3-2-2, according to the yarn reflection signal intensity in the target area, adjust the laser pulse power to , where is the power adjustment factor, is the adjusted target signal strength; S3-2-3, according to the movement speed of the spinning , adjust the laser radar scanning speed to , where is the maximum scanning speed of the laser radar, is the maximum movement speed of the yarn; S3-2-4, generating confidence adjustment information according to adjusting the laser pulse wavelength, adjusting the laser pulse power and adjusting the laser radar scanning speed: .

[0030] By dynamically adjusting the scanning parameters of the lidar according to the real-time evaluation results, the stability and accuracy of the acquired image data are ensured, thereby improving the image quality. The laser power and wavelength are dynamically adjusted so that even when the reflection characteristics of the yarn change, the laser reflection intensity remains within an acceptable range, avoiding image quality problems caused by too low or too high reflection intensity. The movement speed and reflectivity of the yarn in the production environment may change continuously. Dynamic adjustment of the scanning parameters can effectively and promptly respond to such changes, ensuring the consistency and high quality of image acquisition. Different types of yarns may have different reflectivities. Dynamic adjustment of the laser power optimizes the reception of reflected light signals, making the change in reflection intensity more significant when a wire break occurs, thereby improving the accuracy of wire break identification. The movement speeds of the yarns are different, and the adjustment of the scanning speed helps to capture the movement state of the yarn, reduce motion blur, and improve the temporal resolution of the image, ensuring the exact timing of capturing the wire break.

[0031] Step S4 includes the following: S4-1, using filtering to remove background noise in the high confidence image, using histogram equalization to enhance the contrast in the high confidence image, and standardizing the grayscale value of the high confidence image; S4-2, using an edge detection operator to calculate the gradient of the preprocessed high confidence image, detecting the contour features of the yarn in the target area, and obtaining the reflection intensity features according to the average reflection intensity of the yarn contour position; The method for obtaining the profile characteristics and reflection intensity characteristics of the yarn is: S4-2-1, use edge detection operator to calculate the horizontal gradient of the image and the gradient in the vertical direction , the gradient amplitude is calculated as and gradient direction , perform non-maximum suppression on each pixel on the image, and retain strong and weak edges by setting high and low thresholds to obtain edge features of high-confidence images; S4-2-2, define the yarn edge area using a sliding window based on the obtained edge features , calculate the yarn edge area The average reflection intensity of all pixels in the image is calculated as , where is the reflection intensity of all pixels in the yarn edge area, Yarn edge area The total number of pixels.

[0032] Edge is one of the most significant structural features in an image. It highlights the shape and outline of the yarn. In yarn break detection, the edge of the yarn can effectively distinguish the yarn from the background, and in the case of yarn breakage, clearly identify the missing parts at both ends of the yarn. The edge operator is used to effectively extract the outline of the yarn and the potential broken part. After extracting the edge features, the convolutional neural network can more easily learn the location of the broken area and make accurate judgments on the broken yarn. The sliding window method extracts the reflection intensity characteristics of the yarn by calculating the average reflection intensity of each local area. The reflection intensity of the yarn will change during the laser scanning process due to the physical properties of the yarn, such as material, thickness, surface state, etc. Therefore, the reflection intensity is used as an important feature to distinguish the state of normal yarn and broken yarn. The reflection intensity of the yarn changes dramatically when the yarn is broken. The broken area cannot reflect the laser, resulting in low reflection intensity in this part. The sliding window method effectively captures the change in reflection intensity and determines the break position by the change in reflection intensity. The reflection intensity is smoothed by the sliding window method, which effectively reduces the interference of local noise on feature extraction, makes the features more stable and reliable, and enhances the robustness of the convolutional neural network model under different lighting and angles.

[0033] S4-3, establish a convolutional neural network model and use the extracted contour features and reflection intensity features for training, generate a labeled data set based on the known high-confidence image data, each data sample in the labeled data set includes input features and broken line labels, the input features include contour features and reflection intensity features, and the broken line labels are divided into "broken line" and "not broken line" as the output of the convolutional neural network model; S4-4, concatenate the contour feature and the reflection intensity feature into a vector As the input of the convolutional neural network model, the vector The local features are extracted by the convolution layer of the convolutional neural network model, flattened and output into a one-dimensional vector, which enters the fully connected layer for classification, and the disconnection label is output through the output layer. The activation function of the fully connected layer is ReLU, and the activation function of the output layer is sigmoid. S4-5, define the loss function as ,in, is the predicted output of the convolutional neural network model, The convolutional neural network model parameters are optimized by minimizing the loss function, and the model probability value is output through the sigmoid activation function. , where is the output of the last layer of the convolutional neural network; S4-6, define the output threshold as , the model probability value With output threshold For comparison, if the model probability value Greater than or equal to the output threshold , then the image processing server outputs the recognition result through the network as disconnected. If the model probability value Less than the output threshold , then the image processing server outputs the recognition result through the network as being not disconnected; S4-7, after performing disconnection recognition on the high-confidence image transmitted in real time, using the idle computing resources of the image processing server to extract the lidar image with the true image label and the low-confidence image label, and inputting the lidar image data with the true image label and the low-confidence image label into the convolutional neural network model for disconnection recognition, and saving the disconnection recognition results in timestamp order, and deleting all lidar image data with false image labels; S4-8, if a yarn break is identified in the high-confidence image transmitted in real time or the laser radar image with a true image label and a low-confidence image label, the image processing server output device prompts the staff that the yarn is broken.

[0034] Example 2: Please refer to Figure 2 As shown, the present invention is a laser radar image enhanced broken line recognition system for textile materials, including a laser scanning module, a server filtering module, a laser radar image processing module, a feature extraction module, and a deep learning module; The laser scanning module is used to perform laser pulse scanning on the target area, collect the movement change information of the yarn in the target area, and record the scanning mode of the laser pulse, and upload the laser radar image data to the image processing server in real time in combination with the network; the target area is the spinning area from the lower end of the front roller of the spinning frame to the steel ring; The server filtering module is used to receive and process the laser radar image data transmitted by the laser scanning module, specifically including: generating a laser characteristic rate of change according to a scanning pattern of the laser pulse; Generate target feature change rate according to the motion change information of the yarn; Verify the consistency between the laser feature change rate and the target feature change rate, and evaluate the matching between the laser pulse scanning and the yarn movement; The authenticity of the laser radar image is checked according to the matching evaluation result, a confidence model is established, a confidence prior is performed on the laser radar image acquired in real time, and a confidence label is added to the image, wherein the confidence label includes a high confidence image and a low confidence image; The laser radar image processing module is used to perform denoising, enhancement and standardization processing on high-confidence images to generate images to be processed, and extract the reflection intensity fluctuation characteristics in the images to be processed in combination with edge detection methods; A feature extraction module is used to fuse the reflection intensity fluctuation feature with the edge feature and convert it into a feature vector; The deep learning module is used to identify the broken line image using the deep learning method based on the feature vector, and verify the broken line identification result according to the confidence label and authenticity label of the image to be processed.

[0035] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0036] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or may be transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0037] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0038] In the several embodiments provided in the present application, it should be understood that the disclosed systems and devices can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0039] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0040] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0041] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0042] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0043] Finally: The above description is only a preferred 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 identifying broken lines of textile materials using laser radar image enhancement, characterized in that: The specific steps include: Step S1, performing laser pulse scanning on the target area through the laser radar, collecting the yarn movement change information in the target area, and recording the scanning mode of the laser pulse, and uploading the laser radar image data to the image processing server in real time in combination with the network; the target area is the spinning area from the lower end of the front roller of the spinning frame to the steel ring; Step S2, in the image processing server, a laser feature change rate is generated according to the scanning mode of the laser pulse, a target feature change rate is generated according to the yarn motion change information, the laser feature change rate and the target feature change rate are verified for consistency, and the matching of the laser pulse scanning and the yarn motion is evaluated; Step S3, verifying the authenticity of the laser radar image according to the matching evaluation result of the laser pulse scanning and the yarn movement, and establishing a confidence model according to the authenticity test result of the known laser radar image, performing a confidence test on the laser radar image acquired in real time by the image processing server, and adding a confidence label to the laser radar image according to the result of the confidence test, wherein the confidence label includes a high confidence image and a low confidence image; Step S4, denoising, enhancing and standardizing the high confidence image to generate an image to be processed, extracting texture features and reflection intensity fluctuation features in the image to be processed by combining the edge detection method, fusing the texture features and reflection intensity fluctuation features and converting them into feature vectors, using a deep learning method to identify the broken line image, and verifying the broken line identification by combining the confidence label and authenticity label of the image to be processed.

2. The method for identifying broken threads of textile materials using laser radar image enhancement according to claim 1, characterized in that: Step S1 includes the following contents: S1-1, determine the target area of ​​the laser radar, including the location, distribution and quantity of the spun yarn in the target area , the target area is divided into several sub-areas, each of which corresponds to the laser scanning of a spun yarn; S1-2, determine the parameters of the laser radar and the characteristic information of the yarn, including the wavelength range of the laser radar , power range , speed range , the reflectivity of each yarn , the speed of each spun yarn ,in is the index of the spun yarn; S1-3, during the laser radar scanning process, the real-time wavelength and power of the laser radar are recorded as the scanning mode, the reflectivity and movement speed of the yarn in the target area are recorded through laser scanning as the yarn movement change information, the scanning mode and the yarn movement information are timestamped, and the real-time acquired laser radar image and the scanning mode and yarn movement information with the corresponding timestamp are uploaded to the image processing server through the configured network.

3. The method for identifying broken threads of textile materials using laser radar image enhancement according to claim 2, characterized in that: Step S2 includes the following contents: S2-1, according to the scanning mode of the laser pulse, the wavelength of the marking laser pulse when irradiating the previous sub-area is , the laser pulses are separated by a time interval The wavelength when the next sub-area is illuminated is , then the laser pulses are separated by a time interval The wavelength change rate after ; S2-2, according to the scanning mode of the laser pulse, the power of the marking laser pulse when irradiating the previous sub-area is , the laser pulses are separated by a time interval The power when irradiating the next sub-area is , then the laser pulses are separated by a time interval The power change rate after ; S2-3, by defining the influence function of the power and wavelength of the laser scanning on the yarn reflection intensity, the influence degree of the change of the laser adjustment logic on the target area is evaluated to obtain the laser scanning influence coefficient; S2-4, the laser characteristic change rate is obtained according to the wavelength change rate, power change rate and laser scanning influence coefficient of the laser pulse. The calculation expression of the laser characteristic change rate is: ; S2-5, according to the yarn movement change information, the reflectivity and movement speed of each yarn are converted into a time domain signal, and the reflectivity of the yarn at the time is calibrated. for The yarn movement speed is for , and and Perform filtering, denoising and standardization processing; S2-6, use the wavelet decomposition method to decompose the time domain signal after filtering, denoising and standardization, perform wavelet decomposition on the reflectivity of the yarn to obtain the high-frequency coefficients of the reflectivity signal, and calculate the reflectivity change rate , where and Separately for the moment and time The high-frequency coefficient of the reflectivity signal is obtained by wavelet decomposition of the yarn movement speed, and the change rate of the movement speed is calculated as , where and Separately for the moment and time The high frequency coefficient of the motion speed signal; S2-7, obtain the target feature change rate according to the reflectivity change rate and the motion speed change rate. The expression of the target feature change rate is: , where is the target feature change rate, are the weight coefficients of the reflectivity change rate and the motion speed change rate, respectively, and All are positive numbers; S2-8, by calculating the laser characteristic change rate and target feature change rate The Pearson correlation coefficient between the two is used for correlation verification. The calculation expression of the Pearson correlation coefficient is: , where is the laser characteristic change rate and target feature change rate The Pearson correlation coefficient between is the laser characteristic change rate and target feature change rate The covariance between and are the laser characteristic change rates and the target feature change rate The standard deviation of the laser characteristic change rate is calculated by and target feature change rate Euclidean distance to evaluate the laser feature change rate and target feature change rate The difference between , the calculation expression of Euclidean distance is ; S2-9, laser feature change rate according to Pearson correlation coefficient and Euclidean distance and target feature change rate The fit is evaluated, and the fit expression is , set the fit threshold to If the fit Greater than or equal to the fit threshold , then the laser scan of the mark matches the yarn motion characteristics. If the match is Less than the fit threshold , then the marker laser scan deviates from the yarn motion characteristics.

4. The method for identifying broken lines of textile materials using laser radar image enhancement according to claim 3, characterized in that: In S2-3, the influence function of the power and wavelength of the laser scanning on the yarn reflection intensity is defined to evaluate the influence of the laser adjustment logic change on the target area. The logic of obtaining the laser scanning influence coefficient is: S2-3-1, calibrate the wavelength influence function to characterize the influence of the laser wavelength on the target area when the laser adjustment logic changes. Use the quadratic polynomial model to define the wavelength influence function, then the wavelength influence function is , where is the yarn reflection intensity obtained by laser scanning, is the wavelength of the laser scan, where As the model parameters, the quadratic polynomial model is fitted using the least squares method, and the error function is defined as , where For the The actual reflection intensity of the yarn in the scan, is the number of laser scans, is the yarn reflection intensity predicted by the model. about Take the partial derivative and set it to 0, and get the wavelength equation as , solve the wavelength equation to obtain the model parameters , and substitute it into the quadratic polynomial model to obtain the wavelength influence function ; S2-3-2, calibrate the power influence function to characterize the influence of laser power on the target area when the laser adjustment logic changes. Use a logarithmic model to define the power influence function, then the power influence function is , where is the yarn reflection intensity obtained by laser scanning, is the power of laser scanning, where As the model parameters, the least squares method is used to fit the logarithmic model, and the error function is defined as , where For the The actual reflection intensity of the yarn in the scan, is the yarn reflection intensity predicted by the model. about Find the partial derivative and set it to 0, and the power equation is obtained as , solve the power equation to obtain the model parameters , and substitute it into the logarithmic model to obtain the power influence function: ; S2-3-3, the laser scanning influence coefficient is generated based on the wavelength influence function and the power influence function. The expression of the laser scanning influence coefficient is: .

5. The method for identifying broken threads of textile materials using laser radar image enhancement according to claim 4, characterized in that: Step S3 includes the following contents: S3-1, verifying the authenticity of the laser radar image according to the consistency verification. If the laser scan matches the yarn motion feature, a true image label is added to the laser radar image transmitted in real time at the corresponding timestamp. If the laser scan deviates from the yarn motion feature, a false image label is added to the laser radar image transmitted in real time at the corresponding timestamp. S3-2, define the lidar image with the real image label added as 0, and the lidar image with the fake image label added as 1, collect the true and false labels of the lidar image transmitted in real time within the time period T and generate the true and false sequence. The true and false sequence is formed by the combination of 0 and 1 arranged in time order. The number of 0 in the true and false sequence is calibrated as , the total number of 0s and 1s in the true and false sequences is , then the quality coefficient of the true and false sequence is calculated as , the preset quality threshold is , if the quality coefficient Less than or equal to the quality threshold , the output device of the image processing server will prompt the staff to adjust the logic misalignment of the laser scanning and generate confidence adjustment information. If the quality coefficient Greater than quality threshold , the output device of the image processing server prompts the staff to adjust the logic synchronization of the laser scanning, and no confidence adjustment information is generated.

6. The textile material laser radar image enhanced broken thread identification system according to claim 5, characterized in that: The method for verifying the authenticity of the laser radar image in step S3 is: S3-1-1, based on machine learning methods, a confidence model is established for authenticity verification of LiDAR images. The binary classification model is trained using known historical LiDAR image data. The binary classification model is established using the logistic regression method, and the expression is: , where the degree of fit is the input feature, is the probability that the lidar image is a true image, is the weight of the binary classification model, is the bias term of the binary classification model. For a known historical lidar image data, the loss function is defined as , for all known historical lidar image data, calculate the average value of the loss function and use the gradient descent method to optimize the model parameters until the loss function converges or the binary classification model reaches the maximum number of iterations; S3-1-2, use the trained binary classification model to detect the real-time transmitted LiDAR image. If the value of is greater than or equal to the preset authenticity verification threshold, the true image label is added to the lidar image. If the value of is less than the preset authenticity verification threshold, a false image label is added to the lidar image.

7. The textile material laser radar image enhanced broken thread identification system according to claim 5, characterized in that: The logic for generating the confidence adjustment information in step S3 is: S3-2-1, collect the yarn reflection intensity signal in the target area , according to the yarn reflectivity of the target area , adjust the laser pulse wavelength to ; S3-2-2, according to the yarn reflection signal intensity in the target area, adjust the laser pulse power to , where is the power adjustment factor, is the adjusted target signal strength; S3-2-3, according to the movement speed of the spinning , adjust the laser radar scanning speed to , where is the maximum scanning speed of the laser radar, is the maximum movement speed of the yarn; S3-2-4, generating confidence adjustment information according to adjusting the laser pulse wavelength, adjusting the laser pulse power and adjusting the laser radar scanning speed: .

8. The textile material laser radar image enhanced broken thread identification system according to claim 7, characterized in that: Step S4 includes the following: S4-1, using filtering to remove background noise in the high confidence image, using histogram equalization to enhance the contrast in the high confidence image, and standardizing the grayscale value of the high confidence image; S4-2, using an edge detection operator to calculate the gradient of the preprocessed high confidence image, detecting the contour features of the yarn in the target area, and obtaining the reflection intensity features according to the average reflection intensity of the yarn contour position; S4-3, establish a convolutional neural network model and use the extracted contour features and reflection intensity features for training, generate a labeled data set based on the known high-confidence image data, each data sample in the labeled data set includes input features and broken line labels, the input features include contour features and reflection intensity features, and the broken line labels are divided into "broken line" and "not broken line" as the output of the convolutional neural network model; S4-4, concatenate the contour feature and the reflection intensity feature into a vector As the input of the convolutional neural network model, the vector The local features are extracted by the convolution layer of the convolutional neural network model, flattened and output into a one-dimensional vector, which enters the fully connected layer for classification, and the disconnection label is output through the output layer. The activation function of the fully connected layer is ReLU, and the activation function of the output layer is sigmoid. S4-5, define the loss function as ,in, is the predicted output of the convolutional neural network model, The convolutional neural network model parameters are optimized by minimizing the loss function, and the model probability value is output through the sigmoid activation function. , where is the output of the last layer of the convolutional neural network; S4-6, define the output threshold as , the model probability value With output threshold For comparison, if the model probability value Greater than or equal to the output threshold , then the image processing server outputs the recognition result through the network as disconnected. If the model probability value Less than the output threshold , then the image processing server outputs the recognition result through the network as being not disconnected; S4-7, after performing disconnection recognition on the high-confidence image transmitted in real time, using the idle computing resources of the image processing server to extract the lidar image with the true image label and the low-confidence image label, and inputting the lidar image data with the true image label and the low-confidence image label into the convolutional neural network model for disconnection recognition, and saving the disconnection recognition results in timestamp order, and deleting all lidar image data with false image labels; S4-8, if a yarn break is identified in the high-confidence image transmitted in real time or the laser radar image with a true image label and a low-confidence image label, the image processing server output device prompts the staff that the yarn is broken.

9. The method for identifying broken threads of textile materials using laser radar image enhancement according to claim 8, characterized in that: In step S4, the method for obtaining the profile feature and reflection intensity feature of the yarn is: S4-2-1, use edge detection operator to calculate the horizontal gradient of the image and the gradient in the vertical direction , the gradient amplitude is calculated as and gradient direction , perform non-maximum suppression on each pixel on the image, and retain strong and weak edges by setting high and low thresholds to obtain edge features of high-confidence images; S4-2-2, define the yarn edge area using a sliding window based on the obtained edge features , calculate the yarn edge area The average reflection intensity of all pixels in the image is calculated as , where is the reflection intensity of all pixels in the yarn edge area, Yarn edge area The total number of pixels.

10. A laser radar image enhanced broken line identification system for textile materials, used to implement the identification method according to any one of claims 1 to 7, comprising a laser scanning module, a server filtering module, a laser radar image processing module, a feature extraction module, and a deep learning module; The laser scanning module is used to perform laser pulse scanning on the target area, collect the movement change information of the yarn in the target area, and record the scanning mode of the laser pulse, and upload the laser radar image data to the image processing server in real time in combination with the network; the target area is the spinning area from the lower end of the front roller of the spinning frame to the steel ring; The server filtering module is used to receive and process the laser radar image data transmitted from the laser radar device, specifically including: generating a laser characteristic rate of change according to a scanning pattern of the laser pulse; Generate target feature change rate according to the motion change information of the yarn; Verify the consistency between the laser feature change rate and the target feature change rate, and evaluate the matching between the laser pulse scanning and the yarn movement; The authenticity of the laser radar image is checked according to the matching evaluation result, a confidence model is established, a confidence prior is performed on the laser radar image acquired in real time, and a confidence label is added to the image, wherein the confidence label includes a high confidence image and a low confidence image; The laser radar image processing module is used to perform denoising, enhancement and standardization processing on high-confidence images to generate images to be processed, and extract the reflection intensity fluctuation characteristics in the images to be processed in combination with edge detection methods; A feature extraction module is used to fuse the reflection intensity fluctuation feature with the edge feature and convert it into a feature vector; The deep learning module is used to identify the broken line image using the deep learning method based on the feature vector, and verify the broken line identification result according to the confidence label and authenticity label of the image to be processed.

Images