A ring spinning yarn detection method and system based on image analysis

Abstract

The application discloses a kind of ring spinning yarn breakage detection method and system based on image analysis, for spinning frame detection field, the ring spinning yarn breakage detection method includes the following steps: obtaining yarn image and infrared image of infrared sensor capture yarn;Yarn image is identified and analyzed using pre-constructed local area network, and the preliminary detection report of yarn image is obtained;The temperature distribution of yarn is analyzed, and the abnormal temperature change of yarn is identified;Comprehensive analysis abnormal temperature change of yarn and the preliminary detection report of yarn image, and predict the probability of yarn breakage;Predetermined threshold is set, and it is judged that yarn breakage will occur phenomenon;Early warning information is uploaded to host computer through network output node, and early warning information is finally detected and identified.The application can comprehensively evaluate the condition of yarn from multiple angles by combining temperature data and visual data, which helps to more comprehensively understand the state of yarn, thereby improving the accuracy of predicting yarn breakage probability.

Description

Technical Field

[0001] The present invention relates to the field of spinning frame detection, and in particular to a ring spinning yarn break detection method and system based on image analysis. Background Art

[0002] The spinning frame is the core equipment in the spinning process. Its primary function is to transform semi-finished roving or sliver into fine yarn through drafting, twisting (twisting and twisting the fibers), and winding. Each stage of the spinning process, such as drafting and twisting on the roving frame, fiber combing on the comber, and sliver combining and drafting on the gilling machine, is designed to improve the sliver structure and prepare it for final spinning into fine yarn on the ring spinning frame. The spinning frame is not only the main spinning machine, but its output and quality are also key indicators for evaluating the quality of the entire spinning process.

[0003] During the working process of the spinning frame, due to its complexity of operation, various abnormal situations are prone to occur, which may affect the continuity and efficiency of production. These abnormalities may include mechanical failure, raw material problems, operating errors or environmental factors.

[0004] In traditional yarn production technology, temperature monitoring often relies on contact measurement, which may lead to inaccurate data and the inability to effectively monitor and control tiny temperature changes in the production process. In old technologies, yarn quality inspections are usually delayed, and yarn quality problems such as structural defects, excessive stretching or wear cannot be discovered immediately. The prediction and management of production risks such as yarn breakage are usually not advanced enough to effectively prevent and mitigate these risks. Traditional production technologies may involve complex system configurations and expensive hardware, which increases production costs and operational difficulty. Data management and centralized processing are usually not efficient enough, which limits the effective use of data and centralized monitoring of the production process.

[0005] Currently, no effective solutions have been proposed for the problems in related technologies. Summary of the Invention

[0006] In order to overcome the above problems, the present invention aims to propose a ring spinning yarn breakage detection method and system based on image analysis, with the aim of solving the problem that in traditional yarn production technology, temperature monitoring often relies on contact measurement, which may lead to inaccurate data and cannot effectively monitor and control small temperature changes in the production process. In the old technology, the quality detection of yarn is usually delayed, and yarn quality problems such as structural defects, excessive stretching or wear cannot be discovered in time. The prediction and management of production risks such as yarn breakage are usually not advanced enough, and these risks cannot be effectively prevented and mitigated.

[0007] To this end, the specific technical solutions adopted in the present invention are as follows:

[0008] According to one aspect of the present invention, a method for detecting yarn breakage in ring spinning based on image analysis is provided. The method for detecting yarn breakage in ring spinning based on image analysis comprises the following steps:

[0009] S1, obtaining the yarn image and capturing the infrared image of the yarn with the infrared sensor;

[0010] S2. Using a pre-built local area network to identify and analyze the yarn image, and obtain a preliminary yarn image detection report;

[0011] S3. Use hotspot analysis technology to process the infrared image of the yarn, analyze the temperature distribution of the yarn, and identify abnormal temperature changes of the yarn;

[0012] S4. Comprehensively analyze the abnormal temperature changes of the yarn and the preliminary detection report of the yarn image, and predict the probability of yarn breakage;

[0013] S5. Setting a predetermined threshold value. If the predicted probability of yarn breakage exceeds the predetermined threshold value, it is determined that yarn breakage will occur and an early warning message is immediately issued.

[0014] S6. Upload the warning information to the host computer through the network output node, and perform final detection and identification on the warning information.

[0015] Optionally, using a pre-built local area network to identify and analyze the yarn image and obtain a preliminary yarn image detection report includes the following steps:

[0016] S21, building a local area network between a preset number of adjacent integrated devices, defining the integrated device in the middle as a master, and the remaining integrated devices as slaves;

[0017] S22, obtaining the yarn image in the slave machine in the local area network through the host, and uniformly storing the yarn image in the slave machine and the yarn image in the host in a database of the host;

[0018] S23, preprocessing the yarn image using a built-in processor of the host, and performing geometric correction on the preprocessed yarn image;

[0019] S24, using fast Fourier transform to identify the yarn profile and structural features in the preprocessed yarn image, and performing preliminary analysis on the profile and structure;

[0020] S25. Based on the results of the preliminary analysis, the preliminary image features in the slave and the host are fused, and the fused yarn image is analyzed using a convolutional neural network model to extract the final image features;

[0021] S26, identifying time series changes and abnormal patterns of the yarn structure based on the final image features;

[0022] S27. Generate a preliminary yarn image detection report based on the recognition result.

[0023] Optionally, identifying the yarn profile and structure features in the preliminary image features using fast Fourier transform and performing preliminary analysis on the profile and structure comprises the following steps:

[0024] S241, performing fast Fourier transform on the preprocessed yarn image to convert the preprocessed yarn image from the spatial domain to the frequency domain;

[0025] S242. Analyze the results of the fast Fourier transform in the frequency domain;

[0026] S243. Based on the analysis results in the frequency domain, identifying and marking the contour and structural features of the yarn in the preprocessed yarn image;

[0027] S244. Analyze the yarn profile and structural characteristics to determine the overall shape, structural integrity and defects of the yarn.

[0028] Optionally, based on the results of the preliminary analysis, the preliminary image features in the slave and the host are fused, and a convolutional neural network is used to deeply analyze the fused yarn image. Extracting the final image features includes the following steps:

[0029] S251, obtaining yarn profile and structural features, and fusing the yarn profile and structural features using principal component analysis to obtain fused feature data;

[0030] S252. Optimizing parameters of the pre-trained convolutional neural network model using the fused feature data;

[0031] S253, using the optimized convolutional neural network model to preliminarily extract features of the yarn image;

[0032] S254. Using the output of the convolutional neural network model as the input of the K-nearest neighbor algorithm, classifying the features of the extracted yarn image according to the K-nearest neighbor algorithm, and extracting the final image features.

[0033] Optionally, obtaining the yarn profile and structural features, and fusing the yarn profile and structural features using principal component analysis includes the following steps:

[0034] S2511, obtaining the contour and structural features of the yarn extracted from each image;

[0035] S2512, merging the extracted profile and structural features of the yarn into a feature matrix, wherein each row in the feature matrix represents the profile and structural features of the yarn, and each column represents a feature dimension;

[0036] S2513. Calculate the covariance matrix based on the characteristic matrix;

[0037] S2514, performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors of the profile and structural characteristics of the yarn;

[0038] S2515. According to the size of the eigenvalues, select the eigenvectors corresponding to the largest eigenvalues ​​as principal components;

[0039] S2516. Use principal components to transform the contour and structural features of the yarn to obtain fused feature data.

[0040] Optionally, using the output of the convolutional neural network model as the input of the K-nearest neighbor algorithm, and classifying the features of the extracted yarn image according to the K-nearest neighbor algorithm, and extracting the final image features includes the following steps:

[0041] S2541. Select the output of the penultimate layer from the optimized convolutional neural network model as the feature vector;

[0042] S2542. Construct a K-nearest neighbor classifier and determine the number of neighbors;

[0043] S2543, inputting the feature vector into a K-nearest neighbor classifier;

[0044] S2544. For each feature vector to be classified in the K-nearest neighbor classifier, calculate the distance between the feature vector and all feature vectors in the training set;

[0045] S2545. Find K nearest neighbors for each feature vector to be classified based on the calculated distance;

[0046] S2546. Voting is performed based on the categories of the K nearest neighbors, and a category label is assigned to each sample to be classified to obtain the final image features.

[0047] Optionally, processing the infrared image of the yarn using a hotspot analysis technique, analyzing the temperature distribution of the yarn, and identifying abnormal temperature changes of the yarn includes the following steps:

[0048] S31. performing temperature calibration on the infrared image according to the characteristics of the infrared sensor and environmental conditions;

[0049] S32, separating the yarn area from the background area in the infrared image, and extracting the hot spot area of ​​the yarn;

[0050] S33, perform image enhancement on the hotspot area to make the temperature difference obvious;

[0051] S34, converting the image-enhanced infrared image into a temperature distribution map, and displaying the temperature distribution on the yarn;

[0052] S35, analyzing temperature changes on the yarn in the temperature distribution diagram;

[0053] S36. According to the analysis result of the temperature distribution, a temperature distribution pattern is obtained, and abnormal temperature changes on the yarn are identified.

[0054] Optionally, comprehensively analyzing the abnormal temperature change of the yarn and the preliminary detection report of the yarn image and predicting the probability of yarn breakage includes the following steps:

[0055] S41, obtaining a characteristic vector of temperature data in abnormal temperature change of the yarn and a characteristic vector of visual data in a preliminary inspection report of the yarn image;

[0056] S42, calculating the similarity between the feature vector of the temperature data and the feature vector of the visual data using a similarity calculation method;

[0057] S43. Integrate the feature vectors of the temperature data and the feature vectors of the visual data into a data set based on the calculated similarity, and build a collaborative recommendation model based on the integrated data set;

[0058] S44. Optimize the collaborative recommendation model using historical temperature data and historical visual data of yarn;

[0059] S45. Predict the yarn breakage probability of each yarn based on the optimized collaborative recommendation model.

[0060] Optionally, the similarity calculation formula is:

[0061]

[0062] Where sim(A, B) represents the similarity between the feature vector A of the temperature data and the feature vector B of the visual data in the dataset;

[0063] A represents the eigenvector of temperature data;

[0064] B represents the feature vector of visual data;

[0065] A i The i-th element in the feature vector A representing the temperature data;

[0066] B i The i-th element in the feature vector B representing the visual data;

[0067] n represents the number of elements in the dataset.

[0068] According to another aspect of the present invention, there is also provided a ring spinning yarn break detection system based on image analysis, the system comprising: an image acquisition module, an image recognition and analysis module, a hot spot analysis module, a primary analysis and recognition module, an early warning judgment module and a final analysis and recognition module;

[0069] An image acquisition module is used to obtain yarn images and infrared images of the yarn captured by an infrared sensor;

[0070] Image recognition and analysis module, used to identify and analyze yarn images using a pre-built local area network to obtain a preliminary yarn image detection report;

[0071] Hot spot analysis module, used to process infrared images of yarn using hot spot analysis technology, analyze the temperature distribution of yarn, and identify abnormal temperature changes of yarn;

[0072] The initial analysis and recognition module is used to comprehensively analyze the abnormal temperature changes of the yarn and the preliminary detection report of the yarn image, and predict the probability of yarn breakage;

[0073] The early warning judgment module is used to set a predetermined threshold. If the predicted probability of yarn breakage exceeds the predetermined threshold, it is determined that yarn breakage will occur and an early warning message is immediately issued;

[0074] The final analysis and identification module is used to upload the warning information to the host computer through the network output node and perform final detection and identification on the warning information.

[0075] Compared with the existing technology, this application has the following beneficial effects:

[0076] 1. The present invention can efficiently collect and transmit yarn image data by constructing a dedicated local area network and utilizing the configuration of a low-light all-in-one machine. The application of fast Fourier transform and convolutional neural network improves the accuracy of image analysis, so that the contour and structural features of the yarn can be accurately identified and analyzed. A centralized processing method is adopted, that is, all auxiliary low-light all-in-one machines send data to the main low-light all-in-one machine for processing, which reduces the complexity and cost of the system. At the same time, the use of optimized processing flow reduces the dependence on expensive hardware. Through data fusion and deep learning technology, the subtle features of the yarn are fully captured, and the monitoring ability of the yarn quality is improved. The convolutional neural network model is particularly suitable for processing image data and can effectively identify complex features in yarn images. The use of K-nearest neighbor algorithm and principal component analysis method further optimizes the classification and identification of features, improves the ability and accuracy of predicting yarn quality changes, and generates a preliminary inspection report based on detailed image analysis to provide key information for the production line, helping to timely discover and prevent potential production problems.

[0077] 2. The present invention captures the temperature information of the yarn through an infrared sensor, and can accurately monitor the temperature changes of the yarn during the production process. By analyzing the temperature distribution diagram of the yarn, hot spots or abnormal temperature changes, such as excessive stretching, wear or other types of damage, can be quickly identified. Infrared imaging technology for temperature monitoring is a non-contact method that avoids physical interference or damage to the yarn, maintains the continuity of the production process and the integrity of the yarn, and performs image enhancement on the hot spot area to make the temperature difference more obvious, facilitate analysis and identification, and increase the accuracy and efficiency of detection.

[0078] 3. By combining temperature data and visual data, the present invention can comprehensively evaluate the condition of the yarn from multiple angles, which helps to more comprehensively understand the status of the yarn, thereby improving the accuracy of predicting the probability of yarn breakage. The use of similarity calculation method to compare different types of data feature vectors helps to reveal the correlation between temperature changes and visual features, thereby improving the accuracy and reliability of the prediction model. By constructing a collaborative recommendation model and combining historical data with current data, the probability of yarn breakage can be more accurately predicted. The integration of historical and real-time data provides a more robust basis for decision-making. The use of collaborative recommendation models to predict the risk of yarn breakage can help factory managers and operators make more accurate production adjustments and decisions, thereby improving production efficiency and product quality.

[0079] 4. The present invention sets a threshold value and immediately issues an early warning when the risk of yarn breakage reaches a certain level, which helps operators to respond quickly and reduce production interruptions and yarn losses. Through timely early warning and intervention, unexpected shutdowns and material losses caused by yarn breakage can be significantly reduced, ensuring production efficiency and product quality. By uploading early warning information to the host computer, centralized management and monitoring can be achieved, which facilitates data storage and analysis, making production management more efficient and centralized. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] The above characteristics, features and advantages of the present invention and their implementation methods and methods will become more clearly understood in conjunction with the following description of the embodiments, which will be described in detail in conjunction with the accompanying drawings. Here, a schematic diagram is shown:

[0081] Figure 1 is a flow chart of a method for detecting yarn breakage in ring spinning based on image analysis according to an embodiment of the present invention;

[0082] Figure 2 The present invention is a block diagram of a yarn break detection system for ring spinning based on image analysis according to an embodiment of the present invention.

[0083] In the picture:

[0084] 1. Image acquisition module; 2. Image recognition and analysis module; 3. Hotspot analysis module; 4. Initial analysis and recognition module; 5. Early warning judgment module; 6. Final analysis and recognition module. DETAILED DESCRIPTION

[0085] In order to help those skilled in the art better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 those skilled in the art without creative work are within the scope of protection of this application.

[0086] According to an embodiment of the present invention, a method and system for detecting yarn breakage in ring spinning based on image analysis are provided.

[0087] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to one embodiment of the present invention, a method for detecting yarn breakage in ring spinning based on image analysis is provided, and the method for detecting yarn breakage in ring spinning based on image analysis comprises the following steps:

[0088] S1. Acquiring a yarn image and capturing an infrared image of the yarn with an infrared sensor.

[0089] It should be noted that the present invention uses a low-light integrated machine to capture yarn images, while an infrared sensor is used to capture infrared images of the yarn. The low-light integrated machine provides detailed visual information about the yarn's physical appearance, while the infrared sensor provides information about the yarn's thermal properties. This detailed visual information and the yarn's thermal properties will be used in subsequent steps to analyze the yarn's condition and predict potential yarn breakage.

[0090] S2. Use the pre-built local area network to identify and analyze the yarn image and obtain a preliminary detection report of the yarn image.

[0091] Preferably, using a pre-built local area network to identify and analyze the yarn image and obtain a preliminary yarn image detection report includes the following steps:

[0092] S21, building a local area network between a preset number of adjacent integrated devices, defining the integrated device in the middle as a master, and the remaining integrated devices as slaves;

[0093] S22, obtaining the yarn image in the slave machine in the local area network through the host, and uniformly storing the yarn image in the slave machine and the yarn image in the host in a database of the host;

[0094] S23, preprocessing the yarn image using a built-in processor of the host, and performing geometric correction on the preprocessed yarn image;

[0095] S24, using fast Fourier transform to identify the yarn profile and structural features in the preprocessed yarn image, and performing preliminary analysis on the profile and structure;

[0096] S25. Based on the results of the preliminary analysis, the preliminary image features in the slave and the host are fused, and the fused yarn image is analyzed using a convolutional neural network model to extract the final image features;

[0097] S26. Identify time series changes and abnormal patterns in the yarn structure based on the final image features (time series analysis typically involves detecting trends, periodic changes, or sudden changes in the data, looking for subtle changes in the yarn structure that may indicate quality degradation or impending yarn breakage. This is performed with the help of existing machine learning algorithms, which can quickly process large amounts of data and identify subtle changes that may be overlooked by the human eye. When abnormal patterns are detected, such as sudden changes in structure or irregular changes in texture, these changes are marked and an alert is generated. The production line operator can take quick action, such as adjusting production parameters or performing repairs, to prevent potential production problems);

[0098] S27. Generate a preliminary yarn image detection report based on the recognition result.

[0099] In addition, it should be noted that a local area network is established between a series of adjacent low-light all-in-one machines. In the local area network, the machine located in the center is set as the main low-light all-in-one machine, and the surrounding machines are used as auxiliary low-light all-in-one machines. If the preset number is 5, the auxiliary low-light all-in-one machines are sorted from left to right, and the first auxiliary low-light all-in-one machine is assigned the number 1. And so on, each auxiliary low-light all-in-one machine is assigned its own number, so that when processing in the main low-light all-in-one machine, according to the number, it can be specifically found that the low-light all-in-one machine has a problem, and multiple low-light all-in-one machines use one processor, which can greatly save costs.

[0100] The main low-light all-in-one machine has a built-in processing server that is responsible for receiving yarn images from the auxiliary machines and itself. This centralized data reception and processing ensures data consistency and synchronous processing. The received images are first pre-processed, including geometric correction, to ensure that images taken from different angles can be correctly aligned and compared.

[0101] First, the Fast Fourier Transform (FFT) technique is used to analyze the profile and structural features of the yarn. Fast Fourier Transform technology is known for its efficiency and accuracy, and can quickly identify complex patterns and structures in images, which is crucial for understanding the physical properties of the yarn. Subsequently, by fusing the image features from different cameras and applying a convolutional neural network (CNN) model for in-depth analysis, the subtle features of the yarn can be captured more comprehensively. The convolutional neural network model is particularly suitable for processing image data. It provides strong support for identifying the quality of the yarn by learning and extracting complex features in the image. By analyzing the features, the changes in the yarn structure over time and its abnormal patterns are monitored, which is crucial for the early detection of potential production problems. Finally, based on the comprehensive analysis, a preliminary inspection report on the yarn status is generated, which not only reflects the current yarn quality, but also predicts potential risks and problems, thereby providing key information for the manufacturing process.

[0102] Preferably, identifying the yarn profile and structure features in the preliminary image features using fast Fourier transform and performing preliminary analysis on the profile and structure comprises the following steps:

[0103] S241, performing fast Fourier transform on the preprocessed yarn image to convert the preprocessed yarn image from the spatial domain to the frequency domain;

[0104] S242. Analyze the results of the fast Fourier transform in the frequency domain;

[0105] In addition, it should be noted that spectrum analysis: the fast Fourier transform is used to convert the yarn graph data from the spatial domain (or time domain) to the frequency domain. The result of this step is a complex array containing amplitude and phase information; the amplitude of each frequency component (that is, the modulus of the complex number) is calculated and visualized as a spectrum graph. The spectrum graph can reveal the main frequency components in the yarn graph, and the frequency components are related to the pre-processed yarn image; in the spectrum graph, specific peaks or patterns correspond to specific structural features of the yarn, such as texture, periodic patterns, etc. Identifying characteristic frequency components helps to understand the structure of the yarn and possible defects.

[0106] Based on the results of frequency domain analysis, combined with the knowledge of the physical properties of the yarn and the manufacturing process, the fast Fourier transform results are analyzed. By comparing the fast Fourier transform results of normal yarn and defective yarn, typical defect characteristics are discovered and then used for automatic defect detection and quality control.

[0107] S243. Based on the analysis results in the frequency domain, identifying and marking the contour and structural features of the yarn in the preprocessed yarn image;

[0108] S244. Analyze the yarn profile and structural characteristics to determine the overall shape, structural integrity and defects of the yarn.

[0109] Preferably, based on the results of the preliminary analysis, the preliminary image features in the slave and the host are fused, and the fused yarn image is deeply analyzed using a convolutional neural network. Extracting the final image features includes the following steps:

[0110] S251, obtaining yarn profile and structural features, and fusing the yarn profile and structural features using principal component analysis to obtain fused feature data;

[0111] S252. Optimizing parameters of the pre-trained convolutional neural network model using the fused feature data;

[0112] It should be explained that a "pre-trained convolutional neural network model" refers to a neural network model that has been trained on a large amount of data and has learned certain common features. It is usually trained on a large-scale dataset to learn the basic features of data such as images, text, or sound, and can then be further fine-tuned or optimized for specific tasks. In the context of yarn image processing, the pre-trained model can help quickly and effectively extract the key features of the yarn. The present invention uses a public convolutional neural network model, which includes a convolutional layer, a pooling layer, and a fully connected layer.

[0113] Among them, the convolutional layer is the core of the convolutional neural network, responsible for extracting the features of the input yarn image. Through the convolution operation, this layer can capture the local dependencies and spatial hierarchical structures of the image, such as edges and corners.

[0114] The pooling layer usually follows the convolutional layer to reduce the spatial size of the feature map, reduce the computational burden, and simplify the information by taking the maximum value (max pooling) or the average value (average pooling) within the region, which helps to extract a wider range of features and reduce overfitting;

[0115] The fully connected layer is located at the end of the network. Its task is to perform final classification or regression based on the features extracted and combined by the previous layers. Each neuron is connected to all the activation values ​​of the previous layer, so it can integrate the information of the entire yarn image.

[0116] S253, using the optimized convolutional neural network model to preliminarily extract features of the yarn image;

[0117] S254. Using the output of the convolutional neural network model as the input of the K-nearest neighbor algorithm, classifying the features of the extracted yarn image according to the K-nearest neighbor algorithm, and extracting the final image features.

[0118] Preferably, obtaining the yarn profile and structural features, and fusing the yarn profile and structural features using principal component analysis comprises the following steps:

[0119] S2511, obtaining the contour and structural features of the yarn extracted from each image;

[0120] S2512, merging the extracted profile and structural features of the yarn into a feature matrix, wherein each row in the feature matrix represents the profile and structural features of the yarn, and each column represents a feature dimension;

[0121] S2513. Calculate the covariance matrix based on the characteristic matrix;

[0122] S2514, performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors of the profile and structural characteristics of the yarn;

[0123] S2515. According to the size of the eigenvalues, select the eigenvectors corresponding to the largest eigenvalues ​​as principal components;

[0124] S2516. Use principal components to transform the contour and structural features of the yarn to obtain fused feature data.

[0125] It should be explained that the results of the fast Fourier transform are analyzed in the frequency domain to find specific frequency patterns, which are related to the contour and structural characteristics of the yarn. For example, a specific frequency may represent a smooth edge or a complex texture of the yarn; based on the analysis results of the frequency domain, the contour and structural features in the yarn image are identified and marked, and the association between the specific frequency pattern and the specific structure of the yarn is identified. For example, an irregular frequency pattern may indicate a defect in the yarn; analyzing the identified yarn contour and structural features to determine the overall shape, structural integrity and defects of the yarn is the key to evaluating the quality of the yarn, and it can reveal any unevenness or damage to the yarn; the fusion of the extracted yarn contour and structural features is achieved through principal component analysis, which can merge multiple features into a smaller number representative features. In this process, features from different images are first merged, and then the covariance matrix of the feature matrix is ​​calculated and eigenvalue decomposition is performed. By selecting the eigenvector corresponding to the largest eigenvalue as the principal component, we are able to retain the most important information while reducing the dimension of the data. The purpose of using the fused feature data to optimize the parameters of the pre-trained convolutional neural network (CNN) model is to adjust the convolutional neural network model to make it better adapt to the characteristics of the yarn image. The optimized convolutional neural network model is then used to extract the features of the yarn image. The output of the convolutional neural network model is used as the input of the K-nearest neighbor algorithm to classify the extracted yarn image features, thereby extracting the final image features and accurately identifying and classifying specific features in the yarn image, such as different types of yarn structures or defects.

[0126] Preferably, using the output of the convolutional neural network model as the input of the K-nearest neighbor algorithm, and classifying the features of the extracted yarn image according to the K-nearest neighbor algorithm, and extracting the final image features includes the following steps:

[0127] S2541. Select the output of the penultimate layer from the optimized convolutional neural network model as the feature vector;

[0128] S2542. Construct a K-nearest neighbor classifier and determine the number of neighbors;

[0129] S2543, inputting the feature vector into a K-nearest neighbor classifier;

[0130] S2544. For each feature vector to be classified in the K-nearest neighbor classifier, calculate the distance between the feature vector and all feature vectors in the training set;

[0131] S2545. Find K nearest neighbors for each feature vector to be classified based on the calculated distance;

[0132] S2546. Voting is performed based on the categories of the K nearest neighbors, and a category label is assigned to each sample to be classified to obtain the final image features.

[0133] It should be noted that the penultimate layer refers to the layer before the output layer in the convolutional neural network model. In many cases, this layer can capture sufficiently rich and useful features. These features not only contain information about the original data but also have undergone complex transformations by the convolutional neural network model, making them better suited for classification tasks. The K-nearest neighbor algorithm is a distance-based classification method. Its core idea is that the category of a sample can be determined by the categories of several of its nearest neighbors. In this step, the value of "K" (the number of neighbors to consider for each sample) must be determined. The feature vectors extracted by the convolutional neural network model are input into the K-nearest neighbor classifier. The feature vectors represent the key features of the yarn image. The algorithm calculates the distance between each feature vector to be classified and all feature vectors in the training set to determine which samples are nearest neighbors. Based on the calculated distances, the K nearest neighbors are found for each feature vector to be classified. Voting means that the category of each sample to be classified is determined by the category of its K nearest neighbors. The sample to be classified is assigned to the category with the largest number of neighbors.

[0134] S3. Use hotspot analysis technology to process the infrared image of the yarn, analyze the temperature distribution of the yarn, and identify abnormal temperature changes of the yarn.

[0135] Preferably, using hotspot analysis technology to process infrared images of yarns, analyze the temperature distribution of the yarns, and identify abnormal temperature changes of the yarns includes the following steps:

[0136] S31. performing temperature calibration on the infrared image according to the characteristics of the infrared sensor and environmental conditions;

[0137] S32, separating the yarn area from the background area in the infrared image, and extracting the hot spot area of ​​the yarn;

[0138] S33, perform image enhancement on the hotspot area to make the temperature difference obvious;

[0139] S34, converting the image-enhanced infrared image into a temperature distribution map, and displaying the temperature distribution on the yarn;

[0140] S35, analyzing temperature changes on the yarn in the temperature distribution diagram;

[0141] S36. According to the analysis result of the temperature distribution, a temperature distribution pattern is obtained, and abnormal temperature changes on the yarn are identified.

[0142] It should be explained that the infrared image is temperature calibrated according to the characteristics of the infrared sensor and the current environmental conditions, which means adjusting the image to reflect the actual temperature value and eliminating errors caused by environmental changes (such as temperature, humidity) or sensor differences. The yarn area in the infrared image is separated from the background area and the hot spot area of ​​the yarn is extracted. This step is achieved through image segmentation technology. Its purpose is to ensure that the analysis is focused on the yarn rather than other irrelevant areas. The hot spot area is image enhanced to make the temperature difference more obvious. This is achieved by adjusting the image contrast or using a specific image processing algorithm to more clearly identify the high and low temperatures and distribution. The enhanced infrared image is converted into a temperature distribution map and the temperature distribution is displayed on the yarn. The color or brightness value in the image is converted into a corresponding temperature value and presented in an easy-to-understand form. The temperature changes on the yarn in the temperature distribution map are analyzed to identify uneven temperature distribution, local hot spots or other abnormal patterns, which may be indicators of yarn quality problems. Based on the analysis results of the temperature distribution, the temperature distribution pattern is obtained and abnormal temperature changes on the yarn are identified. This is to identify changes that are different from the regular temperature distribution pattern. The changes may indicate certain problems with the yarn, such as excessive stretching, wear or other types of damage.

[0143] S4. Comprehensively analyze the abnormal temperature changes of the yarn and the preliminary detection report of the yarn image, and predict the probability of yarn breakage.

[0144] Preferably, comprehensively analyzing the abnormal temperature change of the yarn and the preliminary detection report of the yarn image and predicting the probability of yarn breakage includes the following steps:

[0145] S41, obtaining a characteristic vector of temperature data in abnormal temperature change of the yarn and a characteristic vector of visual data in a preliminary inspection report of the yarn image;

[0146] S42, calculating the similarity between the feature vector of the temperature data and the feature vector of the visual data using a similarity calculation method;

[0147] S43. Integrate the feature vectors of the temperature data and the feature vectors of the visual data into a data set based on the calculated similarity, and build a collaborative recommendation model based on the integrated data set;

[0148] It should be noted that the similarity calculation helps identify and emphasize the correlation between temperature data and visual data, combining the feature vectors of temperature and visual data into a single dataset for building a collaborative recommendation model. By revealing the correlation between these two data types, this helps improve the accuracy and effectiveness of the prediction model. To a certain extent, the higher the similarity, the more reliable the feature vectors of temperature data and visual data can be inferred. Therefore, the reliability of the integrated dataset can be achieved by using existing techniques such as thresholding methods, which will not be discussed in detail here.

[0149] S44. Optimize the collaborative recommendation model using historical temperature data and historical visual data of yarn;

[0150] S45. Predict the yarn breakage probability of each yarn based on the optimized collaborative recommendation model.

[0151] Preferably, the similarity calculation formula is:

[0152]

[0153] Where sim(A, B) represents the similarity between the feature vector A of the temperature data and the feature vector B of the visual data in the dataset;

[0154] A represents the eigenvector of temperature data;

[0155] B represents the feature vector of visual data;

[0156] A i The i-th element in the feature vector A representing the temperature data;

[0157] B i The i-th element in the feature vector B representing the visual data;

[0158] n represents the number of elements in the dataset.

[0159] It is important to explain that feature vectors are first extracted from the yarn's abnormal temperature change data and yarn image inspection reports. Here, the feature vectors of the temperature data may contain information such as the temperature distribution pattern and extreme points, while the feature vectors of the visual data may include visual features such as the yarn's color, texture, and shape. A similarity calculation method is then used to analyze the similarity between the feature vectors of the temperature data and the feature vectors of the visual data to determine whether there is any overlap or correlation in the patterns of these two different types of data. The feature vectors of the temperature and visual data are then integrated into a single dataset, and a collaborative recommendation model is constructed based on this integrated dataset. The collaborative recommendation model aims to identify the correlation between temperature and visual features to more accurately predict the probability of yarn breakage. The collaborative recommendation model is optimized using historical yarn temperature and visual data. By analyzing this historical data, the parameters of the collaborative recommendation model can be adjusted to make it more suitable for the actual production environment and data characteristics. The optimized collaborative recommendation model then predicts the probability of yarn breakage for each individual yarn. This step combines information from different data sources to obtain a more comprehensive perspective on potential yarn defects and quality issues.

[0160] S5. A predetermined threshold is set. If the predicted probability of yarn breakage exceeds the predetermined threshold, it is determined that yarn breakage will occur and an early warning message is immediately issued.

[0161] It should be explained that first, various data on the production line, such as yarn tension and speed, are collected through sensors and other equipment, and these data are pre-processed. Then, a prediction model is trained using a machine learning algorithm so that it can predict the probability of yarn breakage based on real-time data. Then, an appropriate threshold is set. When the predicted probability exceeds this threshold, it is determined to be a high-risk state. Finally, this model is run in real time during the production process, and once a potential yarn breakage risk is detected (when the location of the yarn breakage risk is identified, the host or slave number is obtained from the local area network based on the local area network where the location is located, and the specific location is determined based on the specific number, so that targeted measures can be taken), an early warning is immediately issued to the operator through the alarm system so that timely measures can be taken to reduce unexpected downtime and material loss, and ensure production efficiency and product quality.

[0162] S6. Upload the warning information to the host computer through the network output node, and perform final detection and identification on the warning information.

[0163] It needs to be explained that the early warning information is sent from the control system of the production line to the host computer via the network output node. This process involves data encoding, encryption and transmission to ensure the security and accuracy of the information. Then, the host computer (mainly responsible for making secondary judgments to improve the accuracy of the alarm) receives the information and conducts summary analysis to realize data storage, processing and decision support. The host computer analyzes the early warning information, performs final detection and identification on it to determine whether it is a real alarm or a false alarm, and responds accordingly, including issuing instructions to the production line, adjusting the production process, or notifying relevant personnel to conduct inspections and repairs, quickly responding to and effectively handling potential problems, and improving the safety and efficiency of the production line.

[0164] According to another embodiment of the present invention, Figure 2 As shown, a ring spinning yarn break detection system based on image analysis is also provided, which includes: an image acquisition module 1, an image recognition and analysis module 2, a hot spot analysis module 3, a primary analysis and recognition module 4, an early warning judgment module 5 and a final analysis and recognition module 6;

[0165] Image acquisition module 1, used to obtain yarn images and infrared images captured by infrared sensors;

[0166] Image recognition and analysis module 2, used to use the pre-built local area network to identify and analyze the yarn image and obtain a preliminary detection report of the yarn image;

[0167] Hot spot analysis module 3, used to process the infrared image of the yarn using hot spot analysis technology, analyze the temperature distribution of the yarn, and identify abnormal temperature changes of the yarn;

[0168] The initial analysis and recognition module 4 is used to comprehensively analyze the abnormal temperature changes of the yarn and the preliminary detection report of the yarn image, and predict the probability of yarn breakage;

[0169] The early warning judgment module 5 is used to set a predetermined threshold value. If the predicted probability of yarn breakage exceeds the predetermined threshold value, it is determined that yarn breakage will occur and an early warning message is immediately issued;

[0170] The final analysis and identification module 6 is used to upload the warning information to the host computer through the network output node and perform final detection and identification on the warning information.

[0171] Specifically, in order to facilitate better understanding by those skilled in the art, the following technical terms or some nouns that may be involved in the present application are explained in the following embodiments:

[0172] Fast Fourier transform is an efficient algorithm used to calculate discrete Fourier transform and its inverse transform. It is a mathematical tool widely used in signal processing, image processing, data analysis and other fields. It converts signals from time domain to frequency domain and reveals the frequency components of the signal.

[0173] The collaborative recommendation model is neural collaborative filtering, which is a recommendation model that combines traditional collaborative filtering with deep learning technology. It mainly learns and predicts user preferences for items by building deep neural networks. The key to this model is that it can capture the complex and nonlinear relationship between users and items through multi-layer neural networks, thereby providing more accurate and personalized recommendations. Neural collaborative filtering can not only process traditional user-item interaction data, but also integrate other auxiliary information, such as user demographic data and item description information. Its advantages lie in its strong nonlinear modeling capabilities, flexible network architecture design, and the ability to effectively deal with the "cold start" problem of new users or new items.

[0174] In summary, with the help of the above technical solutions of the present invention, the present invention can efficiently collect and transmit yarn image data by constructing a dedicated local area network and utilizing the configuration of the low-light all-in-one machine. The application of fast Fourier transform and convolutional neural network improves the accuracy of image analysis, so that the contour and structural features of the yarn can be accurately identified and analyzed. A centralized processing method is adopted, that is, all auxiliary low-light all-in-one machines send data to the main low-light all-in-one machine for processing, which reduces the complexity and cost of the system. At the same time, the use of optimized processing flow reduces the dependence on expensive hardware. Through data fusion and deep learning technology, the subtle features of the yarn are fully captured, and the monitoring ability of the yarn quality is improved. The convolutional neural network model It is particularly suitable for processing image data and can effectively identify complex features in yarn images. The K-nearest neighbor algorithm and principal component analysis method are used to further optimize the classification and identification of features, improve the ability and accuracy of predicting yarn quality changes, and generate preliminary inspection reports based on detailed image analysis to provide key information for the production line, helping to timely discover and prevent potential production problems. The present invention captures the temperature information of the yarn through an infrared sensor, which can accurately monitor the temperature changes of the yarn during the production process. By analyzing the temperature distribution map of the yarn, hot spots or abnormal temperature changes such as over-stretching, wear or other types of damage can be quickly identified. Infrared imaging technology for temperature monitoring is a non-contact method that avoids the risk of damage to the yarn. The physical interference or damage to the yarn is prevented, which maintains the continuity of the production process and the integrity of the yarn. The image of the hot spot area is enhanced to make the temperature difference more obvious, which is convenient for analysis and identification, and increases the accuracy and efficiency of detection. The present invention combines temperature data and visual data to comprehensively evaluate the condition of the yarn from multiple angles, which helps to understand the state of the yarn more comprehensively, thereby improving the accuracy of predicting the probability of yarn breakage. The use of similarity calculation method to compare different types of data feature vectors helps to reveal the correlation between temperature changes and visual features, thereby improving the accuracy and reliability of the prediction model. By constructing a collaborative recommendation model and combining historical data with current data, the probability of yarn breakage can be predicted more accurately. , integrating historical and real-time data to provide a more robust basis for decision-making, and using a collaborative recommendation model to predict the risk of yarn breakage, which can help factory managers and operators make more accurate production adjustments and decisions, thereby improving production efficiency and product quality; the present invention sets a threshold and immediately issues an early warning when the risk of yarn breakage reaches a certain level, which helps operators to respond quickly and reduce production interruptions and yarn losses. Through timely early warning and intervention, it can significantly reduce unexpected shutdowns and material losses caused by yarn breakage, ensuring production efficiency and product quality. By uploading the early warning information to the host computer, centralized management and monitoring can be achieved, which facilitates data storage and analysis, making production management more efficient and centralized.

[0175] Although the present invention has been disclosed above with reference to preferred embodiments, the embodiments are merely examples for the purpose of illustration and are not intended to limit the present invention. Those skilled in the art may make various modifications and alterations without departing from the spirit and scope of the present invention. The scope of protection claimed by the present invention shall be subject to the claims.

Claims

1. A method for detecting yarn breakage in ring spinning based on image analysis, characterized in that: include: S1. Acquire the yarn image and capture the infrared image of the yarn with the infrared sensor; S2. Use the pre-built local area network to identify and analyze the yarn image and obtain a preliminary detection report of the yarn image; S3. Use hotspot analysis technology to process the infrared image of the yarn, analyze the temperature distribution of the yarn, and identify abnormal temperature changes of the yarn; S4. Comprehensively analyze the abnormal temperature changes of the yarn and the preliminary yarn image detection report, and predict the probability of yarn breakage; S5. Set a predetermined threshold. If the predicted probability of yarn breakage exceeds the predetermined threshold, it is determined that yarn breakage will occur and an early warning message is immediately issued; S6. Upload the early warning message to the host computer through the network output node, and finally detect and identify the early warning message; Among them, S2 includes: S21, building a local area network between a preset number of adjacent integrated machines, and defining the integrated machine in the middle position as the host, and the remaining integrated machines as slaves; S22, obtaining the yarn image in the slave machine in the local area network through the host, and uniformly storing the yarn image in the slave machine and the yarn image in the host in the database of the host; S23, using the built-in processor of the host to preprocess the yarn image, and performing geometric correction on the preprocessed yarn image; S24, using fast Fourier transform to identify the yarn contour and structural features in the preprocessed yarn image, and performing preliminary analysis on the contour and structure; S25, based on the results of the preliminary analysis, data fusion of the preliminary image features in the slave machine and the host, and using the convolutional neural network model to analyze the fused yarn image to extract the final image features; S26, identifying the time series changes and abnormal patterns of the yarn structure based on the final image features; S27, generating a preliminary detection report of the yarn image based on the recognition results; Among them, S25 includes: S251, obtaining the yarn profile and structural features, using the principal component analysis method to fuse the yarn profile and structural features to obtain fused feature data; S252, using the fused feature data to optimize the parameters of the pre-trained convolutional neural network model; S253, using the optimized convolutional neural network model to preliminarily extract the features of the yarn image; S254, using the output of the convolutional neural network model as the input of the K-nearest neighbor algorithm, and classifying the extracted yarn image features according to the K-nearest neighbor algorithm, and extracting the final image features.

2. The method for detecting yarn breakage in ring spinning based on image analysis according to claim 1, characterized in that: The method of using fast Fourier transform to identify the yarn profile and structural features in the preliminary image features and performing preliminary analysis on the profile and structure includes the following steps: S241, performing fast Fourier transform on the preprocessed yarn image to convert the preprocessed yarn image from the spatial domain to the frequency domain; S242. Analyze the results of fast Fourier transform in the frequency domain; S243. Based on the analysis results in the frequency domain, identifying and marking the contour and structural features of the yarn in the preprocessed yarn image; S244. Analyze the yarn profile and structural characteristics to determine the overall shape, structural integrity and defects of the yarn.

3. The method for detecting yarn breakage in ring spinning based on image analysis according to claim 2, characterized in that: The method of obtaining the yarn profile and structural features and fusing the yarn profile and structural features using the principal component analysis method comprises the following steps: S2511, obtaining the contour and structural features of the yarn extracted from each image; S2512, merging the extracted profile and structural features of the yarn into a feature matrix, wherein each row in the feature matrix represents the profile and structural features of the yarn, and each column represents a feature dimension; S2513. Calculate the covariance matrix based on the characteristic matrix; S2514, performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors of the profile and structural characteristics of the yarn; S2515. According to the size of the eigenvalues, select the eigenvectors corresponding to the largest eigenvalues ​​as principal components; S2516. Use principal components to transform the contour and structural features of the yarn to obtain fused feature data.

4. The method for detecting yarn breakage in ring spinning based on image analysis according to claim 3, characterized in that: The method of using the output of the convolutional neural network model as the input of the K-nearest neighbor algorithm, classifying the features of the extracted yarn image according to the K-nearest neighbor algorithm, and extracting the final image features includes the following steps: S2541. Select the output of the penultimate layer from the optimized convolutional neural network model as the feature vector; S2542. Construct a K-nearest neighbor classifier and determine the number of neighbors; S2543, inputting the feature vector into a K-nearest neighbor classifier; S2544. For each feature vector to be classified in the K-nearest neighbor classifier, calculate the distance between the feature vector and all feature vectors in the training set; S2545. Find K nearest neighbors for each feature vector to be classified based on the calculated distance; S2546. Voting is performed based on the categories of the K nearest neighbors, and a category label is assigned to each sample to be classified to obtain the final image features.

5. The method for detecting yarn breakage in ring spinning based on image analysis according to claim 4, characterized in that: The method of processing the infrared image of the yarn using the hotspot analysis technology, analyzing the temperature distribution of the yarn, and identifying abnormal temperature changes of the yarn includes the following steps: S31. performing temperature calibration on the infrared image according to the characteristics of the infrared sensor and environmental conditions; S32, separating the yarn area from the background area in the infrared image, and extracting the hot spot area of ​​the yarn; S33, perform image enhancement on the hotspot area to make the temperature difference obvious; S34, converting the image-enhanced infrared image into a temperature distribution map, and displaying the temperature distribution on the yarn; S35, analyzing temperature changes on the yarn in the temperature distribution diagram; S36. According to the analysis result of the temperature distribution, a temperature distribution pattern is obtained, and abnormal temperature changes on the yarn are identified.

6. The method for detecting yarn breakage in ring spinning based on image analysis according to claim 5, characterized in that: The comprehensive analysis of abnormal temperature changes of the yarn and the preliminary detection report of the yarn image and the prediction of the probability of yarn breakage includes the following steps: S41, obtaining a characteristic vector of temperature data in abnormal temperature change of the yarn and a characteristic vector of visual data in a preliminary inspection report of the yarn image; S42, calculating the similarity between the feature vector of the temperature data and the feature vector of the visual data using a similarity calculation method; S43. Integrate the feature vectors of the temperature data and the feature vectors of the visual data into a data set based on the calculated similarity, and build a collaborative recommendation model based on the integrated data set; S44. Optimize the collaborative recommendation model using historical temperature data and historical visual data of yarn; S45. Predict the yarn breakage probability of each yarn based on the optimized collaborative recommendation model.

7. The method for detecting yarn breakage in ring spinning based on image analysis according to claim 6, characterized in that: The calculation formula of the similarity is: Where sim(A, B) represents the similarity between the feature vector A of the temperature data and the feature vector B of the visual data in the dataset; A represents the eigenvector of temperature data; B represents the feature vector of visual data; A i The i-th element in the feature vector A representing the temperature data; B i The i-th element in the feature vector B representing the visual data; n represents the number of elements in the dataset.

8. A ring spinning yarn break detection system based on image analysis, used to implement the ring spinning yarn break detection method based on image analysis according to any one of claims 1 to 7, characterized in that: The system includes: image acquisition module, image recognition and analysis module, hot spot analysis module, initial analysis and recognition module, early warning judgment module and final analysis and recognition module; An image acquisition module is used to obtain yarn images and infrared images of the yarn captured by an infrared sensor; Image recognition and analysis module, used to identify and analyze yarn images using a pre-built local area network to obtain a preliminary yarn image detection report; Hot spot analysis module, used to process infrared images of yarn using hot spot analysis technology, analyze the temperature distribution of yarn, and identify abnormal temperature changes of yarn; The initial analysis and recognition module is used to comprehensively analyze the abnormal temperature changes of the yarn and the preliminary detection report of the yarn image, and predict the probability of yarn breakage; The early warning judgment module is used to set a predetermined threshold. If the predicted probability of yarn breakage exceeds the predetermined threshold, it is determined that yarn breakage will occur and an early warning message is immediately issued; The final analysis and identification module is used to upload the warning information to the host computer through the network output node and perform final detection and identification on the warning information.

Images