Yarn quality intelligent identification system and method

Abstract

The invention relates to the technical field of yarn quality detection, and discloses an intelligent yarn quality recognition system and method.The system comprises a data capture layer, an image optimization layer, a model analysis layer and a user interaction layer; and through optimization processing of noise reduction, color space conversion, image segmentation and the like, a yarn image area needing to be analyzed is extracted. And constructing a defect detection model by adopting a deep learning algorithm, and carrying out defect identification and classification on the yarn image. And the user interaction layer is responsible for displaying a detection result and receiving a user operation instruction to carry out corresponding system operation or parameter adjustment. According to the method, the accuracy and efficiency of yarn quality detection are improved, the defect detection model is constructed by adopting the deep learning algorithm, whether the yarn image has defects and the specific types of the defects can be objectively and consistently judged, and the problem of detection inconsistency caused by subjective judgment in manual detection is avoided.

Description

Technical Field

[0001] The present invention relates to the field of yarn quality detection, and in particular to a yarn quality intelligent identification system and method. Background Art

[0002] As the basic material in the textile industry, the quality of yarn directly affects the performance and quality of the final product.

[0003] Traditional yarn quality inspection mostly relies on manual visual inspection, which is not only time-consuming and labor-intensive, but also easily affected by the subjective judgment of inspectors, resulting in low inspection efficiency and poor consistency. In addition, for yarns on high-speed production lines, manual inspection is even more difficult to keep up with the production rhythm. Although some automated inspection equipment has been applied to the field of yarn quality inspection, these devices can often only identify some more obvious defects, such as large defects or oil stains, and it is difficult to accurately identify minor defects such as one more or one less warp or weft yarn. These minor defects can also have a significant impact on product quality, but existing technologies often cannot effectively detect them. To this end, we propose a yarn quality intelligent identification system and method. Summary of the invention

[0004] The purpose of the present invention is to provide a yarn quality intelligent identification system and method to solve the problems raised in the above background technology.

[0005] To achieve the above-mentioned object, the present invention provides the following technical solutions: a yarn quality intelligent identification system, comprising a data capture layer, an image optimization layer, a model analysis layer and a user interaction layer; The data capture layer includes a visual sensing module and an image transmission module; The visual sensing module uses a high-resolution industrial camera as a core component and is responsible for accurately capturing detailed images of the yarn surface; the image transmission module is responsible for transmitting the captured yarn image data to the image optimization layer; The image optimization layer includes a noise filtering module, a color space conversion module and an image segmentation module; The noise filtering module uses image filtering technology to perform noise reduction on the captured yarn image; the color space conversion module is responsible for converting the denoised color image into a grayscale image; the image segmentation module segments the grayscale image to remove the background or other irrelevant parts, and extracts the yarn image area that needs to be analyzed; The model analysis layer includes a data collection module, a data enhancement processing module, a data synthesis module, a model construction module and a defect identification module; The data collection module is responsible for comprehensively collecting yarn image data, covering normal samples and various defective samples; the data enhancement processing module increases the diversity and quantity of data sets by performing rotation, flipping and cropping transformation operations on the yarn images collected by the data collection module; the model construction module uses a deep learning algorithm to construct a defect detection model, and the defect detection model uses the yarn image as input to determine whether the yarn image has defects and classify the specific categories of defects; The model building module collects the image data output by the data collection module and the data enhancement processing module and integrates them into a rich training data set for training and optimizing the defect detection model; the defect recognition module is used to output the image data output by the image optimization layer to the defect detection model and receive the detection results output by the defect detection model; The user interaction layer includes a result display module and an interactive operation module; The result display module is used to display the detection results output by the defect recognition module, including whether there are defects in the yarn image and the specific category of the defects; the interactive operation module is used to receive operation instructions input by the user and perform corresponding system operations or parameter adjustments according to the operation instructions.

[0006] Preferably, the noise filtering module uses Gaussian filtering to eliminate noise in the image, and replaces the value of each pixel in the image with the weighted average of the pixel values ​​in its neighborhood. The calculation formula is: Among them, G(x,y) is the Gaussian function value at the two-dimensional coordinate (x,y) of the image, x and y are coordinates in the two-dimensional space, σ is the standard deviation of the Gaussian distribution, e is the base of the natural logarithm, and π is the pi. The denoised image is obtained by convolving the Gaussian kernel with the original image.

[0007] Preferably, the image segmentation module uses a threshold segmentation method to extract the yarn image area to be analyzed; first, a threshold T is set, and then the image is segmented into foreground and background parts according to the comparison result between the gray value of the pixel and the threshold. The calculation formula is: in, refer to , f(x,y) is the segmented image, Gray(x,y) is the pixel value of the original grayscale image, T is the set threshold, and through the above formula, the background or other irrelevant parts are segmented and removed, leaving only the image area of ​​the yarn part.

[0008] Preferably, the defect detection model identifies types of yarn defects including thick and thin spots, uneven yarn length, flying fibers, neps, abnormal defects and uneven twist.

[0009] Preferably, the step of training the defect detection model by the model building module includes: Step 1: Data preparation: Aggregate the image data output by the data collection module, data enhancement module, and data synthesis module into a training data set; accurately annotate the collected yarn image data to indicate the location and type of defects, including coarse and fine details, uneven yarn length, flying flowers, cotton neps, abnormal defects, yarn hairiness, and uneven twist; divide the annotated image data into a training set and a validation set, where the training set is used to train the deep learning model, and the validation set is used to evaluate the performance of the model; Step 2: Build a deep learning model: Use a convolutional neural network to build a defect classification model. The model contains multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layer is responsible for extracting features from the image, the pooling layer is used to reduce the dimension and complexity of the features, and the fully connected layer maps the learned features to the final classification results. After the construction is completed, start to initialize the weights and bias parameters of the model. Step 3: Train the deep learning model: Use the training set to train the model; input the image data into the model through the forward propagation algorithm and calculate the output of each layer; use the set cross entropy loss function to calculate the difference between the model prediction and the true annotation; use the back propagation algorithm and gradient descent optimizer to update the model's weights and biases according to the gradient information of the loss function; iterate the training multiple times until the model converges or reaches the preset number of training rounds; Step 4: Evaluate and optimize the model: Use the validation set to evaluate the trained model; calculate the model's accuracy, recall, and F1 score evaluation indicators; optimize the model based on the evaluation results, including adjusting the network structure and modifying hyperparameters.

[0010] Preferably, during the model training process, the mean square error MSE algorithm is used to evaluate the performance of the defect detection model. The specific algorithm is: Where n is the number of data points, is the model prediction value, is the actual value; the prediction accuracy of the model is evaluated according to the MSE value. The smaller the MSE value, the higher the prediction accuracy of the model.

[0011] Preferably, the result display module receives the detection result data output by the defect recognition module, the data including whether there are defects in the yarn image and the specific category information of the defects; further analyzes the detection result data to extract the defect status and defect category of the yarn image; The defect detection results of the yarn image are graphically displayed on the user interactive interface, including highlighting the defect location, marking the defect category and displaying the defect degree.

[0012] Preferably, the interactive operation module provides interface operation functions for the detection results, including zooming in on the defect area, viewing defect details and exporting the detection results.

[0013] Preferably, a method for intelligently identifying yarn quality comprises the following steps: S1. Collecting yarn image data: using an image acquisition device to obtain the yarn image data to be detected; S2. Preprocessing yarn image data: preprocessing the collected yarn image data, including denoising, enhancement and segmentation operations; S3, extracting yarn image features: using image processing technology to extract features of the pre-processed yarn image, including color, texture and shape feature information; S4. Constructing a yarn defect classification model: using deep learning technology to construct a yarn defect classification model, the model is used to receive the extracted yarn image features as input and output the category of the yarn defect; S5. Identifying yarn defects: inputting the extracted yarn image features into a yarn defect classification model to identify the yarn defects and obtain category information of the yarn defects; S6. Outputting yarn quality identification results: displaying the category information of yarn defects to the user in graphical or textual form.

[0014] Compared with the prior art, the present invention has the following beneficial effects: By accurately capturing yarn surface images with high-resolution industrial cameras and combining them with advanced image processing technology, the system can quickly and accurately detect yarn quality, significantly improving detection efficiency and accuracy compared to traditional manual visual inspections.

[0015] The system uses deep learning algorithms to build a defect detection model, which can objectively and consistently determine whether there are defects in the yarn image and the specific category of the defects, avoiding the problem of detection inconsistency caused by subjective judgment in manual inspection. The system integrates data collection and data enhancement processing modules, which can comprehensively collect yarn image data and increase the diversity and quantity of data sets through rotation, flipping and cropping transformation operations, providing rich data support for the training and optimization of defect detection models. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is the overall structure diagram of the present invention; Figure 2 This is the training flow chart of the defect detection model; Figure 3 The figure is a step diagram of the intelligent identification method of yarn quality. DETAILED DESCRIPTION

[0017] 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.

[0018] See also Figure 1-Figure 3 The present invention provides a technical solution: a yarn quality intelligent identification system, including a data capture layer, an image optimization layer, a model analysis layer and a user interaction layer.

[0019] The data capture layer is responsible for capturing the raw image data of the yarn. This layer is mainly composed of a visual sensing module and an image transmission module. The core component of the visual sensing module is a high-resolution industrial camera, which can accurately capture detailed images of the yarn surface to ensure the accuracy of subsequent analysis. The image transmission module is responsible for quickly and stably transmitting these captured yarn image data to the image optimization layer for further processing.

[0020] The main task of the image optimization layer is to preprocess the captured yarn images to improve the image quality and simplify the subsequent analysis process. This layer includes a noise filtering module, a color space conversion module, and an image segmentation module. The noise filtering module uses advanced image filtering technology to effectively remove noise interference in the yarn image. The color space conversion module converts the denoised color image into a grayscale image to reduce the complexity of subsequent processing. The image segmentation module uses an accurate segmentation algorithm to remove the background or other irrelevant parts from the image and extract the yarn image area that needs to be analyzed.

[0021] The model analysis layer is the core of the system and is responsible for realizing the intelligent identification of yarn defects. This layer includes data collection module, data enhancement processing module, data synthesis module, model construction module and defect recognition module. The data collection module comprehensively collects yarn image data, covering normal samples and various defect samples, providing rich basic data for model training. The data enhancement processing module significantly increases the diversity and quantity of data sets and improves the generalization ability of the model by performing transformation operations such as rotation, flipping and cropping on the collected yarn images. The model construction module uses a deep learning algorithm to build a defect detection model, which can accurately determine whether there are defects and the specific category of defects based on the input yarn image. The defect recognition module is responsible for passing the image data output by the image optimization layer into the defect detection model and receiving the detection results output by the model.

[0022] The user interaction layer is responsible for displaying the test results and receiving the user's operation instructions. This layer includes a result display module and an interactive operation module. The result display module displays the test results output by the defect recognition module in an intuitive way, including whether there are defects in the yarn image and the specific category of the defects, so that users can quickly understand the quality status of the yarn. The interactive operation module is responsible for receiving the operation instructions input by the user and performing corresponding system operations or parameter adjustments according to the instructions to ensure the flexibility and ease of use of the system.

[0023] The invention will be further described below in conjunction with Examples 1 to 3: Embodiment 1: The noise filtering module uses Gaussian filtering to eliminate noise in the image. The specific steps are as follows: Step 1: Image acquisition and conversion First, the yarn image to be identified is obtained and converted into a matrix form represented by two-dimensional coordinates (x, y) for subsequent processing.

[0024] Step 2: Gaussian filter design The image is filtered using a Gaussian filter. The two-dimensional function expression of the Gaussian filter is: Where G(x,y) is the value of the Gaussian function at the two-dimensional coordinate (x,y) of the image, x and y are coordinates in two-dimensional space, σ is the standard deviation of the Gaussian distribution, e is the base of the natural logarithm, and π is pi.

[0025] Step 3: Pixel value replacement and weighted average Replace the value of each pixel in the image with the weighted average of the pixel values ​​in its neighborhood. The weighted average is determined by the Gaussian function value G(x, y) designed in step 2, ensuring that the weighted average of the center pixel of the neighborhood is the largest and gradually decreases with increasing distance.

[0026] Step 4: Image convolution operation The Gaussian kernel designed in step 3 is convolved with the original image. During the convolution process, the Gaussian kernel slides on the image, and the weighted sum of the pixel values ​​in its neighborhood is calculated for each pixel position to obtain the denoised image.

[0027] Step 5: Parameter Optimization In practical applications, the Gaussian filter parameter σ is adjusted as needed to achieve the best noise reduction effect.

[0028] The image segmentation module uses a threshold segmentation method to extract the yarn image area that needs to be analyzed; the specific implementation steps are: Step 1: Threshold setting A threshold T is set according to the specific characteristics of the yarn image and experimental experience.

[0029] Step 2: Grayscale value comparison The grayscale value of each pixel (x, y) in the image is extracted and compared with the threshold T set in step 1. The core of the comparison is to determine whether the grayscale value of the pixel reaches or exceeds the threshold T, and use this as a basis to distinguish the foreground and background of the image.

[0030] Step 3: Image Segmentation Based on the comparison result of step 2, the image is segmented. Specifically, if the gray value of a pixel is greater than or equal to the threshold T, the pixel is regarded as the foreground (yarn part) and marked as 1 in the segmented image; conversely, if the gray value is less than the threshold T, it is regarded as the background and marked as 0. This process is achieved through the following formula: Among them, f(x,y) is the segmented image, Gray(x,y) is the pixel value of the original grayscale image, and T is the set threshold.

[0031] Step 4: Remove irrelevant parts After obtaining the preliminary segmented image, in order to further improve the accuracy of the yarn image area, it is necessary to remove the noise points and background areas. The specific methods include: Step 401: remove regions whose area is smaller than a preset threshold, which are often caused by noise or tiny impurities.

[0032] Step 402: Smoothing the edges to eliminate edge noise points and ensure the continuity and integrity of the yarn area.

[0033] Embodiment 2: The steps of training the defect detection model by the model building module include: Step 1: Data preparation The image data output by the data collection module, data enhancement module, and data synthesis module are integrated into a unified training dataset. These data cover various yarn images, including normal yarns and yarns with different types of defects. The diversity and representativeness of the dataset are ensured so that a deep learning model with generalization ability can be trained.

[0034] Each piece of yarn image data collected is annotated in detail to clearly indicate the location and type of defects.

[0035] Defect types include but are not limited to thick and thin spots, uneven yarn evenness, flying fibers, cotton neps, abnormal defects, yarn hairiness, and uneven twist. These annotation information will provide accurate supervision signals for model training.

[0036] The labeled image data is randomly divided into a training set and a validation set, usually divided according to a certain ratio (such as 7:3 or 8:2).

[0037] The training set is used to train the deep learning model to learn the characteristics of yarn images and the recognition rules of defects.

[0038] The validation set is used to evaluate the performance of the model and ensure that the model can show good recognition ability on unseen data.

[0039] Step 2: Build a deep learning model Convolutional neural network (CNN) is used as the basic architecture of the defect classification model. CNN performs well in the field of image processing and can automatically extract high-level features in images.

[0040] The model contains multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layer is responsible for extracting useful features from the image and capturing local dependencies through convolution operations.

[0041] The pooling layer is used to reduce the dimension and complexity of features, improve the computational efficiency of the model, and enhance the spatial invariance of the model.

[0042] The fully connected layer maps the learned features to the final classification results to achieve accurate identification of defects.

[0043] After the model is built, the weight and bias parameters of the model are initialized. The purpose of initialization is to provide a reasonable starting point for the model so that it can converge to the optimal solution faster during training.

[0044] Step 3: Train the deep learning model Use the training set to train the built deep learning model. Before training begins, make sure all training data is prepared and the model parameters are properly initialized.

[0045] The image data is input into the model through the forward propagation algorithm. The image data is processed by the convolution layer, pooling layer and fully connected layer, and finally the model's predicted probability for each defect category is obtained.

[0046] The cross entropy loss function is used to calculate the difference between the model prediction result and the true annotation. The cross entropy loss function can measure the difference between the model prediction probability distribution and the true probability distribution, and is a commonly used loss function for training classification models.

[0047] The gradient of the loss function with respect to the model parameters is calculated through the back propagation algorithm. Based on the gradient information, the gradient descent optimizer is used to update the weight and bias parameters of the model in order to reduce the prediction error in the next iteration. The mean square error MSE algorithm is used to evaluate the performance of the defect detection model. The specific algorithm is: Where n is the number of data points, is the model prediction value, is the actual value; the prediction accuracy of the model is evaluated according to the MSE value. The smaller the MSE value, the higher the prediction accuracy of the model.

[0048] Repeat the above process of forward propagation, loss calculation, back propagation and parameter update until the model converges or reaches the preset number of training rounds. During the training process, the performance of the model on the training set and validation set can be monitored so that the training strategy can be adjusted in time.

[0049] Step 4: Evaluate model performance Use the validation set to evaluate the trained model. Input the image data in the validation set into the model to get the model's prediction results for each image.

[0050] Calculate the model's accuracy, recall, and F1 score, among other evaluation indicators. Accuracy measures the proportion of samples correctly predicted by the model to the total number of samples; recall measures the proportion of defect samples correctly predicted by the model to the actual number of defect samples; and the F1 score is the harmonic mean of accuracy and recall, which is used to comprehensively evaluate the performance of the model.

[0051] The performance of the model is analyzed based on the calculated evaluation indicators. If the performance of the model meets the expected requirements, subsequent optimization and deployment can be carried out; otherwise, the model needs to be further adjusted and optimized.

[0052] Step 5: Optimize model structure According to the performance of the model on the validation set, the structure of the model can be adjusted. For example, the number of convolutional layers can be increased or decreased, the size and step size of the convolution kernel can be adjusted, the type of pooling layer can be changed, etc., to improve the model's feature extraction and generalization capabilities.

[0053] Optimize the model's hyperparameters, including learning rate, batch size, regularization coefficient, etc. By adjusting these hyperparameters, the model's training effect and performance can be further improved.

[0054] After adjusting the model structure and hyperparameters, repeat the training process in step three and the evaluation process in step four until the model performance reaches satisfactory results.

[0055] Embodiment 3: The implementation of the result display module is described in detail below: The result display module first receives the output data from the defect recognition module, which includes whether there are defects in the yarn image and the specific category information of the defects.

[0056] The received test result data is further analyzed to extract the defect status and defect category of the yarn image, ensuring the accuracy and readability of the data.

[0057] On the user interface, the yarn image defect detection results are displayed graphically. Including: Highlight defect location: The specific location of the defect is highlighted on the yarn image through graphical means (such as red frame, circle, etc.), allowing users to identify the defect at a glance.

[0058] Mark defect categories: Mark the defect categories near the defect location or at an appropriate location on the user interface, such as "coarse and fine details", "uneven strips", etc., so that users can understand the specific type of defects.

[0059] Display defect extent: In addition to location and category, you can also display defect extent information, such as defect size, severity, etc., to provide users with a more comprehensive defect assessment.

[0060] User interface design: The user interface should be designed to be intuitive and easy to use, ensuring that users can easily understand and use defect detection results. The interface layout should be reasonable and the graphical elements should be clearly identifiable to improve the user experience.

[0061] The result display module should be able to update the defect detection results in real time and provide instant feedback on the user interface. This ensures that users can obtain the latest detection results at any time and make corresponding processing decisions in a timely manner.

[0062] The implementation methods of the interactive operation module include: The interactive operation module provides users with rich interface operation functions so that they can conduct more in-depth viewing and analysis of the detection results.

[0063] Users can use the magnification tools or gestures on the interface (such as mouse clicks, sliding, etc.) to zoom in on the defect area, allowing users to observe the details of the defect more clearly and make more accurate judgments and analyses.

[0064] When a user clicks or selects a defect, the interactive operation module will display detailed information about the defect, including the defect type, size, location, and possible causes, so that the user can have a more comprehensive understanding of the defect and provide support for subsequent repair and prevention.

[0065] The interactive operation module also provides the function of exporting test results. Users can choose to export the test results in the form of reports or images for further storage, sharing or analysis, meeting the user's usage needs in different scenarios and improving the flexibility and convenience of testing.

[0066] A method for intelligently identifying yarn quality, the method comprising the steps of: S1. Collecting yarn image data: using an image acquisition device to obtain the yarn image data to be detected; S2. Preprocessing yarn image data: preprocessing the collected yarn image data, including denoising, enhancement and segmentation operations; S3, extracting yarn image features: using image processing technology to extract features of the pre-processed yarn image, including color, texture and shape feature information; S4. Constructing a yarn defect classification model: using deep learning technology to construct a yarn defect classification model, the model is used to receive the extracted yarn image features as input and output the category of the yarn defect; S5. Identifying yarn defects: inputting the extracted yarn image features into a yarn defect classification model to identify the yarn defects and obtain category information of the yarn defects; S6. Outputting yarn quality identification results: displaying the category information of yarn defects to the user in graphical or textual form.

[0067] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0068] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A yarn quality intelligent identification system, characterized in that: Includes data capture layer, image optimization layer, model analysis layer and user interaction layer; The data capture layer includes a visual sensing module and an image transmission module; The visual sensing module uses a high-resolution industrial camera as a core component and is responsible for accurately capturing detailed images of the yarn surface; the image transmission module is responsible for transmitting the captured yarn image data to the image optimization layer; The image optimization layer includes a noise filtering module, a color space conversion module and an image segmentation module; The noise filtering module uses image filtering technology to perform noise reduction on the captured yarn image; the color space conversion module is responsible for converting the denoised color image into a grayscale image; the image segmentation module segments the grayscale image to remove the background or other irrelevant parts, and extracts the yarn image area that needs to be analyzed; The model analysis layer includes a data collection module, a data enhancement processing module, a data synthesis module, a model construction module and a defect identification module; The data collection module is responsible for comprehensively collecting yarn image data, covering normal samples and various defective samples; the data enhancement processing module increases the diversity and quantity of data sets by performing rotation, flipping and cropping transformation operations on the yarn images collected by the data collection module; the model construction module uses a deep learning algorithm to construct a defect detection model, and the defect detection model uses the yarn image as input to determine whether the yarn image has defects and classify the specific categories of defects; The model building module collects the image data output by the data collection module and the data enhancement processing module and integrates them into a rich training data set for training and optimizing the defect detection model; The defect recognition module is used to input the image data output by the image optimization layer into the defect detection model, and receive the detection result output by the defect detection model; The user interaction layer includes a result display module and an interactive operation module; The result display module is used to display the detection results output by the defect recognition module, including whether there are defects in the yarn image and the specific category of the defects; the interactive operation module is used to receive operation instructions input by the user and perform corresponding system operations or parameter adjustments according to the operation instructions.

2. The yarn quality intelligent identification system according to claim 1, characterized in that: The noise filtering module uses Gaussian filtering to eliminate noise in the image, replacing the value of each pixel in the image with the weighted average of the pixel values ​​in its neighborhood. The calculation formula is: Among them, G(x,y) is the Gaussian function value at the two-dimensional coordinate (x,y) of the image, x and y are coordinates in the two-dimensional space, σ is the standard deviation of the Gaussian distribution, e is the base of the natural logarithm, and π is the pi. The denoised image is obtained by convolving the Gaussian kernel with the original image.

3. A yarn quality intelligent identification system according to claim 2, characterized in that: The image segmentation module uses a threshold segmentation method to extract the yarn image area that needs to be analyzed; first, a threshold T is set, and then the image is segmented into foreground and background parts according to the comparison result between the gray value of the pixel and the threshold. The calculation formula is: in, refer to , f(x,y) is the segmented image, Gray(x,y) is the pixel value of the original grayscale image, T is the set threshold, and through the above formula, the background or other irrelevant parts are segmented and removed, leaving only the image area of ​​the yarn part.

4. The yarn quality intelligent identification system according to claim 3 is characterized in that: The defect detection model identifies types of yarn defects including thick and thin spots, uneven yarn length, flying fibers, neps, abnormal defects and uneven twist.

5. The yarn quality intelligent identification system according to claim 4, characterized in that: The steps of training the defect detection model by the model building module include: Step 1: Data preparation: Aggregate the image data output by the data collection module, data enhancement module, and data synthesis module into a training data set; accurately annotate the collected yarn image data to indicate the location and type of defects, including coarse and fine details, uneven yarn length, flying flowers, cotton neps, abnormal defects, yarn hairiness, and uneven twist; divide the annotated image data into a training set and a validation set, where the training set is used to train the deep learning model, and the validation set is used to evaluate the performance of the model; Step 2: Build a deep learning model: Use a convolutional neural network to build a defect classification model. The model contains multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layer is responsible for extracting features from the image, the pooling layer is used to reduce the dimension and complexity of the features, and the fully connected layer maps the learned features to the final classification results. After the construction is completed, start to initialize the weights and bias parameters of the model. Step 3: Train the deep learning model: Use the training set to train the model; input the image data into the model through the forward propagation algorithm and calculate the output of each layer; use the set cross entropy loss function to calculate the difference between the model prediction and the true annotation; use the back propagation algorithm and gradient descent optimizer to update the model's weights and biases according to the gradient information of the loss function; iterate the training multiple times until the model converges or reaches the preset number of training rounds; Step 4: Evaluate and optimize the model: Use the validation set to evaluate the trained model; calculate the model's accuracy, recall, and F1 score evaluation indicators; optimize the model based on the evaluation results, including adjusting the network structure and modifying hyperparameters.

6. The yarn quality intelligent identification system according to claim 5, characterized in that: During the model training process, the mean square error (MSE) algorithm is used to evaluate the performance of the defect detection model. The specific algorithm is: Where n is the number of data points, is the model prediction value, is the actual value; the prediction accuracy of the model is evaluated according to the MSE value. The smaller the MSE value, the higher the prediction accuracy of the model.

7. The yarn quality intelligent identification system according to claim 6, characterized in that: The result display module receives the detection result data output by the defect recognition module, which includes whether the yarn image has defects and the specific category information of the defects; further analyzes the detection result data to extract the defect status and defect category of the yarn image; The defect detection results of the yarn image are graphically displayed on the user interactive interface, including highlighting the defect location, marking the defect category and displaying the defect degree.

8. The yarn quality intelligent identification system according to claim 7, characterized in that: The interactive operation module provides interface operation functions for the detection results, including zooming in on the defect area, viewing defect details and exporting the detection results.

9. A method for intelligently identifying yarn quality, characterized in that: The steps of the method include: S1. Collecting yarn image data: using an image acquisition device to obtain the yarn image data to be detected; S2. Preprocessing yarn image data: preprocessing the collected yarn image data, including denoising, enhancement and segmentation operations; S3, extracting yarn image features: using image processing technology to extract features of the pre-processed yarn image, including color, texture and shape feature information; S4. Constructing a yarn defect classification model: using deep learning technology to construct a yarn defect classification model, the model is used to receive the extracted yarn image features as input and output the category of the yarn defect; S5. Identifying yarn defects: inputting the extracted yarn image features into a yarn defect classification model to identify the yarn defects and obtain category information of the yarn defects; S6. Outputting yarn quality identification results: displaying the category information of yarn defects to the user in graphical or textual form.

Images