Intelligent detection method and system for tensile strength of polyester thread

Abstract

The invention provides an intelligent detection method and system for tensile strength of a polyester thread, and relates to the field of quality detection.The method comprises the steps that firstly, multi-scale morphological characteristics of the polyester thread in the stretching process are captured in real time, and local detail characterization of a non-uniform area of the diameter of the polyester thread is enhanced under regulation and control of a dynamic receptive field; and then, mapping the space state characteristic information of the line diameter of the pixel-level polyester thread into line diameter continuous distribution data through a characteristic decoder, synchronously associating a time sequence force value signal of a tension sensor, establishing a cross-modal mapping relationship of image-line diameter-force value, and finally realizing self-adaptive compensation calculation of the tensile strength based on a sectional area dynamic correction model. Coupling interference of wire diameter fluctuation and tension dispersion on strength evaluation is eliminated, dependence of traditional detection on material uniformity assumption is broken through, and a basis is provided for tensile strength detection and performance evaluation of polyester threads.

Description

Technical Field

[0001] This application relates to the field of quality inspection, and more particularly, to an intelligent detection method and system for the tensile strength of polyester threads in the embodiments of this application. Background Art

[0002] As an important industrial textile material, the tensile strength of polyester thread is a key performance index for measuring product quality, directly affecting the safety and reliability of end products such as sewing products, medical suture threads, and industrial lifting belts. In traditional tensile strength detection, after measuring the fracture force value with a tensile force sensor, the tensile strength is calculated through preset wire diameter parameters (strength = force value / cross-sectional area).

[0003] During the production process of polyester thread, due to reasons such as process fluctuations and raw material differences, there are inherent characteristics of uneven axial distribution of wire diameter and large discreteness of initial tension. However, the thickness of the wire diameter directly affects the load-bearing capacity. The strength is high where the wire diameter is thick and low where the wire diameter is thin. If only the force value is used as the strength index during testing while ignoring the wire diameter difference, the test results will not be able to truly reflect the strength characteristics of the material itself, and there will be a lack of comparability between samples with different wire diameters. That is to say, this non-uniformity of physical properties leads to two defects in traditional polyester thread tensile strength detection methods: First, the significant deviation between the preset wire diameter value and the actual cross-sectional area will cause the amplification of strength calculation errors; Second, the strength data between different test samples lose comparability due to wire diameter differences, seriously affecting the objectivity of quality assessment. Therefore, an optimized intelligent detection scheme for the tensile strength of polyester thread is expected. Summary of the Invention

[0004] To solve the above technical problems, this application is proposed. The embodiments of this application provide an intelligent detection method and system for the tensile strength of polyester thread, which first capture the multi-scale morphological characteristics of the polyester thread during the stretching process in real time, and strengthen the local detail representation of the non-uniform area of the polyester thread wire diameter under the regulation of a dynamic receptive field. Then, through a feature decoder, the pixel-level polyester thread wire diameter spatial state feature information is mapped into wire diameter continuous distribution data, synchronously associated with the time-series force value signal of the tensile force sensor, establishing a cross-modal mapping relationship of "image - wire diameter - force value", and finally realizing the adaptive compensation calculation of the tensile strength based on the cross-sectional area dynamic correction model to eliminate the coupling interference of wire diameter fluctuations and tension discreteness on strength assessment, breaking through the dependence of traditional detection on the assumption of material uniformity, and providing a basis for the tensile strength detection and performance evaluation of polyester thread. According to one aspect of this application, an intelligent detection method for the tensile strength of polyester thread is provided, which includes:

[0005] Collecting a tensile detection image of the polyester thread sample object to be detected through a high-resolution camera;

[0006] Identify the region of interest of the polyester thread target and extract features from the tensile test image to obtain the polyester thread diameter state features; perform significant enhancement processing of the polyester thread diameter state features based on the dynamic receptive field to obtain enhanced polyester thread diameter state features;

[0007] Based on the enhanced polyester thread diameter state features, determine the polyester thread diameter data, and calculate the cross-sectional area of the polyester thread according to the polyester thread diameter data;

[0008] Collect the tensile force value data of the polyester thread sample object to be detected through a tensile force sensor;

[0009] Divide the tensile force value data by the cross-sectional area of the polyester thread to obtain the tensile strength data of the polyester thread sample object to be detected. According to another aspect of the present application, there is provided an intelligent detection system for the tensile strength of polyester thread, which includes:

[0010] A polyester thread sample object tensile test image acquisition module for acquiring a tensile test image of the polyester thread sample object to be detected through a high-resolution camera;

[0011] A polyester thread target region of interest identification and feature extraction module for identifying the region of interest of the polyester thread target and extracting features from the tensile test image to obtain the polyester thread diameter state features;

[0012] A polyester thread diameter state feature significant enhancement processing module for performing significant enhancement processing of the polyester thread diameter state features based on the dynamic receptive field to obtain enhanced polyester thread diameter state features;

[0013] A polyester thread cross-sectional area calculation module for determining the polyester thread diameter data based on the enhanced polyester thread diameter state features,

[0014] and calculating the cross-sectional area of the polyester thread according to the polyester thread diameter data;

[0015] A tensile force value data acquisition module for the polyester thread sample object to be detected for collecting the tensile force value data of the polyester thread sample object to be detected through a tensile force sensor;

[0016] A tensile strength data acquisition module for the polyester thread sample object to be detected for dividing the tensile force value data by the cross-sectional area of the polyester thread to obtain the tensile strength data of the polyester thread sample object to be detected.

[0017] Compared with the prior art, a method and system for intelligent detection of the tensile strength of polyester yarn provided by the present application first capture the multi-scale morphological characteristics of the polyester yarn in real time during the stretching process, and enhance the local detail representation of the non-uniform area of the polyester yarn diameter under the regulation of the dynamic receptive field. Then, through the feature decoder, the pixel-level polyester yarn diameter spatial state feature information is mapped into the continuous distribution data of the diameter, and the time-series force value signal of the tensile force sensor is synchronously correlated to establish a cross-modal mapping relationship of "image - diameter - force value". Finally, based on the cross-sectional area dynamic correction model, the adaptive compensation calculation of the tensile strength is realized to eliminate the coupling interference of the diameter fluctuation and the tension discreteness on the strength evaluation, break through the dependence of the traditional detection on the assumption of material uniformity, and provide a basis for the detection and performance evaluation of the tensile strength of polyester yarn. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The above and other objects, features, and advantages of the present application will become more apparent by describing the embodiments of the present application in more detail with reference to the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application, and do not constitute a limitation to the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 It is a flowchart of the method for intelligent detection of the tensile strength of polyester yarn according to an embodiment of the present application.

[0020] Figure 2 It is a schematic diagram of the data flow of the method for intelligent detection of the tensile strength of polyester yarn according to an embodiment of the present application.

[0021] Figure 3 It is a flowchart for identifying and extracting the region of interest of the polyester yarn target and the features from the tensile detection image to obtain the polyester yarn diameter state features in the method for intelligent detection of the tensile strength of polyester yarn according to an embodiment of the present application.

[0022] Figure 4 It is a flowchart for significantly enhancing the polyester yarn diameter state features based on the dynamic receptive field to obtain the enhanced polyester yarn diameter state features in the method for intelligent detection of the tensile strength of polyester yarn according to an embodiment of the present application.

[0023] Figure 5 It is a system block diagram of the system for intelligent detection of the tensile strength of polyester yarn according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] The following will detail various exemplary embodiments, features, and aspects of the present application with reference to the accompanying drawings. The same reference numerals in the drawings denote elements having the same or similar functions. Although various aspects of the embodiments are shown in the drawings, the drawings do not have to be drawn to scale unless otherwise specified.

[0025] As used herein, the term "exemplary" means "serving as an example, embodiment, or illustration". Any embodiment described as "exemplary" herein is not necessarily to be construed as superior to or better than other embodiments.

[0026] In addition, for a better illustration of the present application, numerous specific details are given in the following detailed description. Those skilled in the art should understand that the present application can be implemented without some of these specific details. In some instances, methods, means, elements, and circuits well known to those skilled in the art are not described in detail so as to highlight the gist of the present application.

[0027] Furthermore, the terms "first" and "second" are used for descriptive purposes only and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present application, "a plurality of" means two or more unless specifically defined otherwise.

[0028] As a key industrial textile material, the tensile strength of polyester thread is crucial for ensuring the safety and reliability of end products such as sewing products, medical sutures, and industrial lifting slings. However, due to factors such as process fluctuations and raw material differences during production, the thread diameter of polyester thread often shows uneven distribution along the axial direction, and there is a large discreteness in the initial tension, which makes the thread diameter directly affect the load-bearing capacity and thus affect the strength performance. Traditional tensile strength detection methods measure the fracture force value through a tensile force sensor and calculate the strength based on preset thread diameter parameters. This method increases the strength calculation error due to ignoring the actual changes in the thread diameter, and at the same time, due to the thread diameter differences between different samples, the strength data lacks comparability, seriously affecting the objectivity and accuracy of quality assessment. The traditional tensile strength detection method for polyester thread is difficult to truly reflect the strength characteristics of the material itself.

[0029] Existing improvement techniques for how to eliminate the influence of the physical properties of polyester thread itself (such as uneven thread diameter, initial tension difference, etc.) on the test results mostly focus on mechanical contact diameter measurement or optical projection method. However, mechanical contact diameter measurement causes secondary errors due to the change in pressure during contact diameter measurement that changes the fiber morphology, and the optical projection method is limited by the morphological simplification assumption of two-dimensional projection, and it is difficult to achieve accurate analysis of the thread diameter during the dynamic stretching process.

[0030] To address the above technical problems, in the technical solution of the present application, an intelligent detection method for the tensile strength of polyester thread is proposed. By integrating machine vision and deep learning technologies, a dynamic analysis model for the thread diameter of polyester thread and a force value coupling analysis framework are constructed, breaking through the static assumption of preset thread diameter parameters in traditional detection.

[0031] Specifically, a high-resolution industrial camera is used to capture the multi-scale morphological features of the polyester thread in real time during the stretching process, and the local detail representation of the non-uniform area of the polyester thread diameter is enhanced under the regulation of the dynamic receptive field. Then, through the feature decoder, the pixel-level polyester thread diameter spatial state feature information is mapped into the continuous distribution data of the thread diameter, and the time-series force value signal of the tensile force sensor is synchronously correlated to establish a cross-modal mapping relationship of "image - thread diameter - force value". Finally, based on the cross-sectional area dynamic correction model, the adaptive compensation calculation of the tensile strength is realized, the coupling interference of the thread diameter fluctuation and the tension discreteness on the strength evaluation is eliminated, the dependence on the material uniformity assumption of the traditional detection is broken, and a basis is provided for the tensile strength detection and performance evaluation of the polyester thread.

[0032] This application proposes an intelligent detection method for the tensile strength of polyester threads. Figure 1 It is a flowchart of the intelligent detection method for the tensile strength of polyester threads according to an embodiment of this application. Figure 2 It is a schematic diagram of data flow of the intelligent detection method for the tensile strength of polyester threads according to an embodiment of this application. As Figure 1 and Figure 2 shown, the intelligent detection method for the tensile strength of polyester threads according to an embodiment of this application includes: S110, collecting a tensile detection image of the polyester thread sample object to be detected through a high-resolution camera; S120, identifying and extracting the region of interest of the polyester thread target from the tensile detection image to obtain the polyester thread diameter state feature; S130, performing a significant enhancement process on the polyester thread diameter state feature based on the dynamic receptive field to obtain an enhanced polyester thread diameter state feature; S140, determining the polyester thread diameter data based on the enhanced polyester thread diameter state feature, and calculating the cross-sectional area of the polyester thread according to the polyester thread diameter data; S150, collecting the tensile force value data of the polyester thread sample object to be detected through a tensile force sensor; S160, dividing the tensile force value data by the cross-sectional area of the polyester thread to obtain the tensile strength data of the polyester thread sample object to be detected.

[0033] In the above intelligent detection method for the tensile strength of polyester thread, in step S110, a tensile detection image of the polyester thread sample object to be detected is collected by a high-resolution camera. It should be understood that as an important industrial textile material, the tensile strength of polyester thread is a key performance indicator for measuring product quality, directly affecting the safety and reliability of end products such as sewing products, medical suture threads, and industrial lifting belts. In the actual production process, the wire diameter of the polyester thread often shows uneven distribution along the axial direction, and there is a large discreteness in the initial tension. This non-uniformity of the wire diameter will cause the polyester thread to exhibit different mechanical properties during the tensile process. Traditional tensile strength detection methods often cannot truly reflect the strength characteristics of the material itself because they ignore the actual changes in the wire diameter. During the tensile detection process, the wire diameter of the polyester thread will change with the increase in tensile force, and these changes may be very subtle, especially in areas with a thinner wire diameter. Using a high-resolution camera to collect the tensile detection image of the polyester thread sample object can provide clear and detailed image information, making the measurement of the polyester thread wire diameter more accurate. In contrast, due to limited pixels, a low-resolution camera may not be able to clearly distinguish the details of the polyester thread, especially when the change in the wire diameter is small, which is likely to cause measurement errors. In addition, the tensile detection image of the polyester thread sample object collected by the high-resolution camera can provide more detailed information, which is helpful for subsequent image processing and analysis. For example, during the image denoising and grayscale processing, the high-resolution image can better retain the edge and texture information of the polyester thread, thereby improving the accuracy of identifying the region of interest and extracting features of the polyester thread target.

[0034] Figure 3 It is a flowchart for identifying the region of interest and extracting features of the polyester thread target from the tensile detection image to obtain the wire diameter state characteristics of the polyester thread in the intelligent detection method for the tensile strength of polyester thread according to an embodiment of the present application. As Figure 3 shown, in the embodiment of the present application, step S120, identifying the region of interest and extracting features of the polyester thread target from the tensile detection image to obtain the wire diameter state characteristics of the polyester thread includes: S121, performing image denoising and grayscale processing on the tensile detection image to obtain a grayscale tensile detection image; S122, extracting the region of interest of the polyester thread target from the grayscale tensile detection image to obtain an image of the region of interest of the polyester thread target; S123, extracting the wire diameter state characteristics of the polyester thread from the image of the region of interest of the polyester thread target to obtain the wire diameter state characteristics of the polyester thread.

[0035] Specifically, in step S121, the tensile detection image is subjected to image denoising and grayscale processing to obtain a grayscale tensile detection image. It should be understood that during the tensile strength detection of polyester yarn, due to factors such as the surface reflection characteristics of polyester yarn, uneven illumination in the industrial site, and camera sensor noise, high-frequency noise, speckle interference, and color channel differences often exist in the original tensile detection image. These interferences will cause the target edge to be blurred or artifacts to be generated, directly affecting the accuracy of subsequent wire diameter feature extraction. Therefore, in order to eliminate the interference of external environmental noise and redundant color information on the wire diameter feature analysis, in the technical solution of this application, the tensile detection image is subjected to image denoising and grayscale processing to obtain a grayscale tensile detection image. By using an image denoising algorithm (such as non-local mean filtering or wavelet threshold denoising) to suppress high-frequency noise and adopting the weighted average method to compress the RGB three-channel information into a single grayscale channel, the data dimension can be effectively reduced, the feature representation standard can be unified, and at the same time, the key spatial information of the polyester yarn morphology structure can be retained.

[0036] Specifically, in step S122, the region of interest (ROI) of the polyester yarn target is extracted from the grayscale tensile detection image to obtain an ROI image of the polyester yarn target. It should be understood that due to the existence of complex background elements such as equipment frames, fixture edges, and light reflections in the industrial detection environment, the redundant information in the non-polyester yarn region of the original grayscale image will interfere with the accurate positioning of the wire diameter feature. Especially when the contrast between the polyester yarn edge and the background is low, it is easy to cause mis-segmentation of the polyester yarn body contour. Therefore, in order to eliminate the negative impact of background noise and irrelevant interference on the wire diameter feature analysis, in the technical solution of this application, the region of interest (ROI) of the polyester yarn target is further extracted from the grayscale tensile detection image to obtain an ROI image of the polyester yarn target. By using an ROI extraction algorithm based on edge detection or region growing, it is possible to focus on the main region of the polyester yarn and eliminate background interference, effectively reducing the computational load of the subsequent deep learning model and avoiding non-target region noise being mis-identified as wire diameter features. This move aims to construct a pure wire diameter analysis space, ensuring that the DenseNet feature extractor only performs multi-scale morphology analysis on the polyester yarn body, strengthening the attention weight of the model on the non-uniform region of the wire diameter, and providing a basis for subsequent wire diameter data determination and polyester yarn tensile strength calculation.

[0037] Specifically, in step S123, the polyester thread diameter state feature is extracted from the polyester thread target region of interest image to obtain the polyester thread diameter state feature, including: passing the polyester thread target region of interest image through a polyester thread diameter state feature extractor based on DenseNet to obtain a polyester thread diameter state feature map as the polyester thread diameter state feature. It should be understood that considering the local diameter mutation in the axial distribution of the polyester thread and the microscopic undulation of the surface texture, it is difficult for traditional convolutional neural networks to effectively capture multi-scale spatial features, especially the correlation between fine-grained diameter changes and macroscopic morphology. Therefore, in order to solve the complexity problem of dynamic characterization of diameter non-uniformity, in the technical solution of this application, the polyester thread target region of interest image is further passed through a polyester thread diameter state feature extractor based on DenseNet to obtain a polyester thread diameter state feature map. In particular, the polyester thread diameter state feature extractor based on DenseNet realizes cross-layer reuse of the shallow edge information and deep semantic features of the polyester thread diameter state through dense skip connections, while retaining the high-frequency details of the wire body contour, enhancing the perception ability of the diameter gradient region, and thus overcoming the information attenuation problem in feature transmission of a single-path network. This step aims to construct a multi-level diameter representation with spatial continuity, extract key parameters such as gradient changes, local curvature, and texture anisotropy in the axial distribution of the diameter through a dense feature fusion mechanism, and provide a high-dimensional feature base for subsequent dynamic receptive field enhancement. In practical applications, this polyester thread diameter state feature can not only accurately reflect the spatial distribution pattern of the diameter mutation region (such as the sudden diameter shrinkage caused by spinning process defects), but also establish an implicit correlation between the diameter state and mechanical properties through non-linear interaction between feature channels, enabling the decoder to effectively distinguish real diameter changes from artifacts caused by image noise when reconstructing diameter data, and ultimately supporting the computational robustness of the cross-sectional area dynamic correction model to ensure a high degree of consistency between the tensile strength evaluation result and the material's intrinsic characteristics.

[0038] Figure 4 The figure is a flowchart for significantly enhancing the polyester thread diameter state feature based on a dynamic receptive field in the polyester thread tensile strength intelligent detection method according to an embodiment of the present application to obtain an enhanced polyester thread diameter state feature. As Figure 4As shown, in the embodiment of the present application, in step S130, the polyester thread diameter state feature is subjected to a significant enhancement process of the polyester thread diameter state feature based on a dynamic receptive field to obtain an enhanced polyester thread diameter state feature, including: S131, performing information compression based on feature decoupling on the polyester thread diameter state feature map to obtain a polyester thread diameter state feature vector to be enhanced and distilled; S132, based on the spatial distribution structure of the polyester thread diameter state feature vector to be enhanced and distilled, screening out a set of pixel-level polyester thread diameter state feature vectors within the local receptive field from the set of pixel-level polyester thread diameter state feature vectors; S133, based on the set of pixel-level polyester thread diameter state feature vectors within the local receptive field, performing saliency enhancement on the polyester thread diameter state feature vector to be enhanced to obtain an enhanced polyester thread diameter state feature vector; S134, aggregating multiple enhanced polyester thread diameter state feature vectors to obtain an enhanced polyester thread diameter state feature map as the enhanced polyester thread diameter state feature. It should be understood that since the diameter of the polyester thread undergoes a sudden or gradual dynamic change along the axial direction during the stretching process (such as local sudden contraction, diameter deviation caused by spinning defects), it is difficult for the traditional feature extraction network with a fixed receptive field to balance the contradiction between global shape perception and local detail capture: an overly large receptive field will blur the boundary of the diameter mutation, while an overly small receptive field cannot perceive the overall trend of the gradual change region. Therefore, in order to solve the problem of local adaptability loss in the feature representation of the non-uniform diameter region, in the technical solution of the present application, the polyester thread diameter state feature map is further subjected to a significant enhancement process of the polyester thread diameter state feature based on a dynamic receptive field to obtain an enhanced polyester thread diameter state feature map. Through the significant enhancement process of the polyester thread diameter state feature based on a dynamic receptive field, it is possible to adaptively adjust the receptive field size and shape of each pixel point based on the spatial distribution characteristics of the pixel-level polyester thread diameter state feature vector after feature decoupling (such as texture anisotropy, gradient direction consistency), so that the network can intelligently select the context information aggregation range according to the local mutation degree of the diameter. For example, in the area of sudden diameter contraction, the system automatically reduces the receptive field to focus on high-frequency edge details; while in the area of gentle gradual change, the receptive field is expanded to capture long-range dependencies. The purpose of this process is to establish a spatial adaptive enhancement paradigm for the diameter state feature. After removing redundant noise through feature distillation compression, based on the conformal commutation constraint (meeting the regularity standard) between the volume space representation vector and the boundary representation vector, the collaborative optimization of multi-scale features within the local receptive field is realized.

[0039] In an embodiment of the present application, step S131, performing information compression based on feature decoupling on the polyester thread diameter status feature map to obtain a to-be-strengthened distilled polyester thread diameter status feature vector, includes: S1311, performing feature decoupling on the polyester thread diameter status feature map along the channel dimension to obtain a set of polyester thread diameter status pixel-level feature vectors; S1312, extracting the polyester thread diameter status pixel-level feature vector at the (i, j) pixel position from the set of polyester thread diameter status pixel-level feature vectors as the to-be-strengthened polyester thread diameter status feature vector; S1313, performing information compression on the to-be-strengthened polyester thread diameter status feature vector to obtain the to-be-strengthened distilled polyester thread diameter status feature vector. In particular, the strengthened polyester thread diameter status feature vector is the channel feature vector at the (i, j) pixel position of the strengthened polyester thread diameter status feature map.

[0040] Specifically, step S1311, performing feature decoupling on the polyester thread diameter status feature map along the channel dimension to obtain a set of polyester thread diameter status pixel-level feature vectors, which is expressed by the polyester thread diameter status feature decoupling formula as:

[0041] F ∈ R H×W×0

[0042]

[0043] where F is the polyester thread diameter status feature map, R is the set of real numbers, H, W, and C are the height, width, and number of channels of F respectively, FeatureDecoupling(F) is to perform feature decoupling on F, v 1,1 and v H,Ware the polyester line diameter state pixel-level feature vectors at the (1,1)th and (H,W)th pixel positions in the set of polyester line diameter state pixel-level feature vectors, respectively. It should be understood that in the process of intelligent detection of polyester line tensile strength, the polyester line diameter state feature map usually contains multi-channel feature information, and these channels may imply morphological features of different scales (such as edges, textures, gradients, etc.). However, the direct dependency between channels may lead to feature coupling, so that the information of a specific channel is interfered with or masked by other channels. For example, low-level channels may carry high-frequency edge information, while high-level channels may encode semantic information. If the channels are not decoupled, it is impossible to effectively distinguish the contribution of different channels to the representation of local details of the line diameter. Therefore, this step decouples the polyester line diameter state feature map along the channel dimension to independently parse the multi-channel feature vectors at each pixel position, thereby eliminating redundant associations between channels and providing a basis for subsequent refined processing. By breaking the coupling between channels, pixel-level independent feature representation can be constructed. Specifically, the feature decoupling operation treats each pixel position of the polyester line diameter state feature map as an independent unit, extracts the multi-channel feature vector corresponding to the polyester line diameter state feature map, and forms a set of pixel-level polyester line diameter state pixel-level feature vectors. This process aims to separate the contribution of different channels to the representation of the line diameter state, so that the polyester line diameter state pixel-level feature vector of each pixel can independently reflect the multi-dimensional attributes of the position (such as local line diameter thickness, texture direction, morphological mutation, etc.). By eliminating the mutual interference between channels, feature decoupling provides a more fine-grained feature basis for subsequent dynamic receptive field regulation, so that the model can adaptively enhance the independent features of each pixel. In this way, the obtained set of polyester line diameter state pixel-level feature vectors can more accurately characterize the non-uniformity of the line diameter (such as the detailed differences in the local sudden contraction or expansion area), and the removal of the inter-channel dependency reduces the interference of redundant noise on the feature representation, so that subsequent processing (such as dynamic receptive field adjustment) can adaptively capture the significant area of line diameter change based on the independent feature distribution of each pixel. In addition, pixel-level decoupling also provides flexibility for multi-scale feature fusion, which helps to explore the implicit relationship between wire diameter state and mechanical properties.

[0044] Specifically, step S1312 extracts the polyester line diameter state pixel-level feature vector at the (i, j)th pixel position from the set of polyester line diameter state pixel-level feature vectors as the polyester line diameter state feature vector to be strengthened, and the polyester line diameter state feature acquisition formula to be strengthened is expressed as:

[0045] v tbs =v i,j ∈R C

[0046] Among them, v i,jis the pixel-level feature vector of the polyester thread diameter state at the (i, j) pixel position in the set of pixel-level feature vectors of the polyester thread diameter state, v tbs is the feature vector of the polyester thread diameter state to be enhanced. It should be understood that due to the sudden or gradual change of the thread diameter along the axial direction during the stretching process of the polyester thread, it is difficult for traditional global feature processing methods to capture subtle thread diameter changes. Therefore, it is necessary to extract the pixel-level feature vector of the polyester thread diameter state at a specific position from the set of pixel-level feature vectors of the polyester thread diameter state to construct an independent analysis unit for each pixel. This step stems from the need for refined characterization of the local state of the thread diameter, aiming to avoid the loss of details caused by global feature averaging by focusing on the independent features at the (i, j) pixel position, and to provide an accurate local processing anchor point for subsequent dynamic receptive field enhancement. By extracting the pixel-level feature vector of the polyester thread diameter state at this position as the object to be enhanced, local context associations can be established around it. For example, according to the gradient distribution or texture directionality in the region of sudden thread diameter change, the receptive field range can be adaptively adjusted. This design enhances the network's perception ability of non-uniform regions of the thread diameter, ensuring that local enhancement operations can focus on key pixels, thereby improving the mapping accuracy between the thread diameter state feature and mechanical properties. By taking the pixel-level feature vector of the polyester thread diameter state as the central anchor point, the network can dynamically aggregate multi-scale features in its neighborhood (such as the sharpness of the thread diameter edge, the axial gradual change trend, etc.), eliminating the interference of global noise on local details. This per-pixel processing mechanism effectively improves the sensitivity of the model to regions of sudden thread diameter change (such as sudden diameter shrinkage caused by spinning defects), enabling subsequent dynamic enhancement operations to accurately correct the thread diameter data and supporting the robustness of cross-sectional area calculation.

[0047] Specifically, in step S1313, the information of the feature vector of the polyester thread diameter state to be enhanced is compressed to obtain the feature vector of the distilled polyester thread diameter state to be enhanced, which is represented by the information compression formula of the distilled polyester thread diameter state to be enhanced:

[0048]

[0049] where, ‖·‖ is the first norm of the vector, v sIt is the state feature vector of the diameter of the polyester thread to be enhanced by distillation. It should be understood that the state feature vector of the diameter of the polyester thread to be enhanced usually contains multi-dimensional feature information (such as local texture, edge gradient, morphological curvature, etc.), and there may be redundant or noisy interference in this information (such as light noise during image acquisition or specular artifacts on the surface of the thread). If subsequent processing (such as dynamic receptive field adjustment) is directly based on the original high-dimensional features, it is likely to lead to a sharp increase in computational complexity and difficulty in focusing on key features. Therefore, it is necessary to compress and refine the core representation of the state feature vector of the diameter of the polyester thread to be enhanced by distillation, eliminate the interference of redundant noise on the diameter state analysis, and ensure that subsequent steps can efficiently utilize the information related to the diameter significance. By using the knowledge distillation mechanism to denoise and reduce the dimension of the state feature vector of the diameter of the polyester thread to be enhanced by distillation, a compact and high-information-density state feature vector of the polyester thread to be enhanced by distillation can be generated. Specifically, information compression does not simply discard some features, but rather selects the key components strongly related to the dynamic change of the diameter (such as the gradient direction consistency in the diameter mutation area, local texture anisotropy, etc.) through non-linear transformation (such as norm normalization and directional constraint). This process enables the model to ignore non-significant features (such as background noise or repeated textures in uniform areas), thereby providing a feature basis with high confidence for the size prediction of the dynamic receptive field, and ensuring that subsequent significance enhancement operations focus on real diameter changes rather than artifact interference. In this way, the obtained state feature vector of the polyester thread to be enhanced by distillation significantly reduces dimensional redundancy while retaining the key information of the diameter. This high-purity feature representation not only improves the computational efficiency of dynamic receptive field adjustment but also enhances the sensitivity of the model to non-uniform areas of the diameter. Finally, this step provides a robust feature input for the establishment of the cross-modal mapping relationship (image - diameter - force value), supporting the improvement of the accuracy of tensile strength calculation.

[0050] In the embodiment of the present application, in step S132, based on the spatial distribution structure of the state feature vector of the diameter of the polyester thread to be enhanced by distillation, a set of pixel-level state feature vectors of the diameter of the polyester thread within the local receptive field is selected from the set of pixel-level state feature vectors of the diameter of the polyester thread, including: S1321, based on the characteristics of the feature distribution spatial structure of the state feature vector of the diameter of the polyester thread to be enhanced by distillation, determining the size of the feature receptive field of the state feature vector of the diameter of the polyester thread to be enhanced; S1322, based on the size of the feature receptive field, selecting a set of pixel-level state feature vectors of the diameter of the polyester thread within the local receptive field from the set of pixel-level state feature vectors of the diameter of the polyester thread.

[0051] Specifically, in step S1321, based on the characteristics of the feature distribution spatial structure of the state feature vector of the diameter of the polyester thread to be enhanced by distillation, the size of the feature receptive field of the state feature vector of the diameter of the polyester thread to be enhanced is determined, and it is expressed by the feature receptive field size determination formula as:

[0052]

[0053] where log2 is the logarithmic function value with base 2, and r is the size of the characteristic receptive field of v i,j The non-uniformity of the polyester thread diameter (such as local sudden shrinkage or expansion) results in significant differences in the spatial distribution of the thread diameter state characteristics in different regions. The traditional fixed-size receptive field cannot effectively adapt to this dynamic change. After the information compression of the characteristic vector of the polyester thread diameter state to be enhanced by distillation, its spatial structure characteristics (such as gradient direction consistency, texture anisotropy) already implicitly contain the key information of local sudden changes or gradual changes in the thread diameter. If a preset fixed receptive field size is used, local details will be blurred (such as the boundary of sudden shrinkage of the thread diameter) or long-range dependencies will be missing (such as the trend of the gradual change region) due to ignoring the spatial heterogeneity of the feature distribution. Therefore, it is necessary to dynamically adjust the receptive field size according to the spatial structure characteristics of the characteristic vector of the polyester thread diameter state to be enhanced by distillation to adapt to the analysis requirements of different thread diameter regions and establish a dynamic adaptive local context awareness mechanism. Specifically, in this step, based on the spatial characteristics of the compressed characteristic vector of the polyester thread diameter state to be enhanced by distillation (such as the intensity of local gradient sudden change, texture contrast), the characteristic vector of the polyester thread diameter state to be enhanced by distillation is mapped to a receptive field size parameter through a non-linear function (such as logarithmic transformation). This mechanism makes the receptive field selection semantically guided, ensuring that the local enhancement operation strictly matches the actual physical state of the thread diameter. In this way, a strong correlation is formed between the dynamic receptive field size and the local characteristics of the thread diameter: for the thread diameter mutation region with rich high-frequency details, reducing the receptive field can enhance the edge sharpness and the positioning accuracy of the sudden shrinkage boundary; for the gradual change region dominated by low frequency, expanding the receptive field can effectively model the continuous change trend of the thread diameter axis. This adaptability significantly improves the pertinence of feature enhancement. For example, at the sudden shrinkage of the thread diameter caused by spinning process defects, a small receptive field can accurately capture the micron-level deformation characteristics caused by the sudden change in diameter and avoid the background noise interference introduced by a large receptive field. At the same time, the dynamic size mechanism enhances the generalization ability of the model to complex thread diameter distributions (such as periodic thickness fluctuations) by eliminating the rigid constraints of preset parameters, laying a foundation for the robustness of the cross-modal mapping relationship. In particular, in the technical solution of this application, the receptive field size is no longer preset prior knowledge, but is adaptively determined by the feature content, making the selection of the receptive field more semantic and targeted, more effectively using context information, and improving the accuracy and robustness of saliency detection.

[0054] Specifically, in step S1322, based on the size of the characteristic receptive field, a set of pixel-level polyester thread diameter state characteristic vectors within the local receptive field is screened from the set of pixel-level polyester thread diameter state characteristic vectors, which is represented by the pixel-level polyester thread diameter state screening formula within the local receptive field as:

[0055]

[0056] Among them, W is a set of pixel-level polyester thread diameter state feature vectors within the local receptive field, and v i-r,j , v i+r,j , v m,n , v i,j+r and v i+r,j+r are respectively the pixel-level polyester thread diameter state feature vectors at the (i-r, j), (i+r, j), (m, n), (i, j+r) and (i+r, j+r) pixel positions within the local receptive field in the set of pixel-level polyester thread diameter state feature vectors within the local receptive field. It should be understood that since the dynamic receptive field size has been adaptively determined according to the spatial structure characteristics of the feature distribution (such as small size in the sudden shrinkage area and large size in the gradual change area), it is necessary to screen out the local features in the corresponding spatial range from the set of pixel-level polyester thread diameter state feature vectors based on this size. This operation solves the problems of overloading of context information (large size introduces irrelevant noise) or missing (small size ignores long-range dependencies) caused by the inability of a fixed receptive field to adapt to the non-uniformity of the thread diameter, ensuring that the local enhancement operation only focuses on the thread diameter state features strongly related to the central pixel. By constructing a local context information library centered on the set of pixel-level polyester thread diameter state feature vectors, a spatial correlation feature base can be provided for subsequent saliency enhancement. Specifically, by retrieving the feature vectors of surrounding pixels through a matrix range defined by a dynamic size (such as a square area with a radius r), local information physically related to the thread diameter state of the central pixel can be aggregated (such as edge gradient consistency in the sudden shrinkage area and axial smoothness in the gradual change area). This spatial range screening enables the model to dynamically balance detail capture and trend perception according to the local characteristics of the thread diameter (such as the intensity of texture mutation), for example, enhancing high-frequency edge details within a small receptive field and modeling the axial gradual change law of the thread diameter within a large receptive field. In this way, the set of pixel-level polyester thread diameter state feature vectors within the obtained local receptive field accurately covers the physical context related to the thread diameter state of the central pixel. For example, for the thread diameter sudden shrinkage point caused by spinning defects, the local features screened by the small-size receptive field include the high-gradient features of the sudden shrinkage boundary and the diameter comparison information in the adjacent area, supporting edge sharpening enhancement; for the gentle gradual change area, the large-size receptive field aggregates long-range axial features to reveal the continuity law of the thread diameter change. This dynamic spatial adaptation mechanism significantly improves the semantic consistency of feature enhancement, avoids irrelevant noise introduced by a fixed receptive field (such as distal background texture), enables subsequent weighted fusion to strengthen the thread diameter saliency features according to physical relevance, and provides high-fidelity input data for dynamic cross-sectional area correction.

[0057] In an embodiment of the present application, step S133, based on the set of pixel-level polyester thread diameter state feature vectors within the local receptive field, significantly enhances the polyester thread diameter state feature vector to be enhanced to obtain an enhanced polyester thread diameter state feature vector, including: S1331, performing conformal representation fusion on the set of pixel-level polyester thread diameter state feature vectors within the local receptive field to obtain a pixel-level polyester thread diameter state fusion feature vector within the local receptive field; S1332, performing weighted fusion on the pixel-level polyester thread diameter state fusion feature vector within the local receptive field and the polyester thread diameter state feature vector to be enhanced to obtain an enhanced polyester thread diameter state feature vector.

[0058] Specifically, the processing procedure of step S133 is as follows:

[0059] First, determine the trainable weighting coefficients α and β. It should be understood that for the polyester thread diameter state feature vector v tbs corresponding to the set w of pixel-level polyester thread diameter state feature vectors within the local receptive field, for v m,n (m = i - r ~ i + r, n = j - r ~ j + r), the set W of pixel-level polyester thread diameter state feature vectors within the local receptive field usually contains two different characterization modes: the body space (the uniform region inside the thread diameter) and the boundary (the region of abrupt change or significant gradient of the thread diameter). If the body space features and boundary features are directly fused, feature conflicts may occur due to incompatible spatial representations (for example, the vector direction differences between the low-gradient features in the body space and the high-gradient features at the boundary), thereby weakening the enhancement effect of the significant components. Therefore, it is necessary to establish a commutation constraint between the body space and boundary features through conformal representation fusion, eliminate the representation ambiguity of the two in the vector space, ensure that the spatial topological structure of the fused feature vector strictly corresponds to the physical state of the thread diameter, and achieve conformal alignment of the two types of features in the tensor space. Specifically, by defining the body space vector and the boundary vector and constructing the spatial two-norm representation of the commutation difference vector, the fusion process is forced to follow the conformal invariance of the physical space. Among them, it is necessary to first determine the body space representation vector as:

[0060]

[0061] The boundary representation vector is:

[0062]

[0063] Then, by modulating the trainable weighting coefficients α and β, the surface type-body space tensor is made to have the regularity satisfying the commutation relationship, that is, the body space representation vector v m,n (3) and the boundary representation vector v m,n (2) the spatial two-norm representation of the commutation difference vector between them tends to the product of the coefficients α and β:

[0064] ||v m,n (3) -v m,n (2) ||2 = ω × α × β

[0065] where ω is the equal - ratio scaling coefficient.

[0066] In this way, for the region with uniform wire diameter, the volume - space vector weight is enhanced to maintain smoothness; for the mutation region, the boundary - vector weight is strengthened to amplify the gradient feature. This commutation constraint dynamically adjusts the contribution ratio of the two types of features through the equal - ratio scaling coefficient, so that the pixel - level polyester wire diameter state fusion feature vector within the fused local receptive field maintains the continuity of the overall shape (such as the axial gradient trend) while retaining the local details of the wire diameter (such as the sharpness of the sudden - shrink edge). In this way, the set of pixel - level polyester wire diameter state feature vectors within the local receptive field is fused into a unified expression with spatial conformality. For example, at the boundary of the sudden - shrink of the wire diameter, the boundary vector dominates through the high - gradient - amplitude feature, strengthening the edge contrast and direction consistency of the sudden - shrink region; while in the region with uniform wire diameter, the volume - space vector suppresses noise interference through the low - variance feature, maintaining the smoothness representation of the axial texture. This conformal fusion mechanism enables the enhanced pixel - level polyester wire diameter state fusion feature vector within the local receptive field to accurately distinguish the physical state categories of the wire diameter (mutation / gradual change / uniform) and improve the consistency of cross - scale features. Finally, the pixel - level polyester wire diameter state fusion feature vector within the local receptive field provides a high - fidelity input for dynamic weighted fusion by eliminating the volume - edge feature conflict, significantly reducing the mean square error (MSE) and the local outlier ratio of the tensile strength prediction.

[0067] After that, based on the determined trainable weighting coefficients α and β above, the set of pixel - level polyester wire diameter state feature vectors within the local receptive field is fused, and the pixel - level polyester wire diameter state fusion feature vector within the local receptive field and the polyester wire diameter state feature vector to be enhanced are weighted - fused to obtain the enhanced polyester wire diameter state feature vector. The formula representation of this process is as follows:

[0068]

[0069] where α and β are trainable weighting hyperparameters, θ(v m,n ) is the significant enhancement weight factor of v m,n , softmax(·) is the softmax function, is the vector multiplication, v’ i,j is v i,jThe enhanced polyester thread diameter status feature vector at the (i, j) pixel position in the corresponding enhanced polyester thread diameter status feature map. In the embodiments of the present application, the enhanced polyester thread diameter status feature vector is the channel feature vector at the (i, j) pixel position of the enhanced polyester thread diameter status feature map. In particular, the trainable weighting hyperparameter here should be understood that although the enhanced polyester thread diameter status feature vector already contains the independent features of the central pixel, its isolation may lead to insufficient representation of the local mutation or gradual change trend of the thread diameter (for example, ignoring the edge continuity of the sudden contraction area or the axial smoothness of the gradual change area). The pixel-level polyester thread diameter status fusion feature vector within the local receptive field aggregates the context information physically associated with the central pixel (such as the gradient direction and texture contrast of adjacent pixels). However, directly replacing the original features will lose the detailed specificity of the central position. Therefore, weighted fusion is required to balance the relationship between the central feature and the local context, solve the representation limitation of a single feature source, and ensure that the saliency enhancement process retains both the micro-scale features of the central pixel and incorporates the macroscopic morphological relevance. This step dynamically adjusts the contribution ratio of the central feature and the local context feature through the learnable weight parameter to generate an enhanced polyester thread diameter status feature vector with both local details and global consistency. Specifically, weighted fusion is performed through the determined trainable weighting hyperparameter above, and different weights are assigned to the pixel-level polyester thread diameter status fusion feature vector within the local receptive field and the enhanced polyester thread diameter status feature vector to be enhanced through a soft attention mechanism (such as feature similarity calculation). This mechanism enables the model to adaptively adjust the feature enhancement strategy according to the local characteristics of the thread diameter. For example, it amplifies the context edge information in the spinning defect area and weakens the redundant texture interference in the normal area to achieve the optimal coupling of the central pixel feature and the local context.

[0070] Specifically, in step S134, multiple reinforced polyester thread diameter state feature vectors are subjected to feature aggregation to obtain a reinforced polyester thread diameter state feature map as the reinforced polyester thread diameter state feature. It should be understood that the reinforced polyester thread diameter state feature vectors contain rich information, but this information is often scattered in different feature vectors. Through the aggregation operation, these scattered information can be fused to extract more meaningful features. In the tensile strength detection of polyester threads, each reinforced polyester thread diameter state feature vector contains detailed information about the local thread diameter state, such as the thickness of the thread diameter, the change of texture, etc. Aggregating these reinforced polyester thread diameter state feature vectors can fuse the local thread diameter state information into a global reinforced polyester thread diameter state feature map to extract higher-level features. These higher-level features can better reflect the overall characteristics of the polyester thread and provide a more accurate basis for subsequent strength calculation. Moreover, considering that the representation ability of each reinforced polyester thread diameter state feature vector may be limited, which makes the model's perception ability of the thread diameter state feature of the polyester thread weak. Aggregating these reinforced polyester thread diameter state feature vectors can enhance the representation ability of the reinforced polyester thread diameter state feature, thereby improving the model's perception ability of the thread diameter state feature of the polyester thread and enhancing the robustness to interference factors. Specifically, in a specific embodiment of the present application, first, each reinforced polyester thread diameter state feature vector is regarded as a local feature representation. These reinforced polyester thread diameter state feature vectors are spatially distributed, and each reinforced polyester thread diameter state feature vector corresponds to a specific position in the image. Through the convolution operation, these local reinforced polyester thread diameter state feature vectors are weighted and summed by the convolution kernel to generate a new feature map. Next, the convolution feature map is downsampled through the pooling operation. The pooling operation usually adopts max pooling or average pooling, and its purpose is to retain the most important feature information while reducing the feature dimension. Max pooling will select the maximum value in the local area as the representative, while average pooling will calculate the average value in the local area. After the convolution and pooling operations, the obtained feature map already contains the aggregated feature information. To further enhance the feature representation ability, batch normalization (BatchNormalization) and activation functions (such as ReLU) are introduced. Batch normalization can normalize the feature map, making the distribution of features more stable, thereby improving the training efficiency and performance of the model. The activation function can introduce non-linear factors, enabling the model to learn more complex feature relationships. Finally, a global reinforced polyester thread diameter state feature map representation is obtained.In an embodiment of the present application, step S140, based on the state characteristics of the diameter of the reinforced polyester thread, determines the diameter data of the polyester thread, and calculates the cross-sectional area of the polyester thread according to the diameter data of the polyester thread, including: S141, passing the state characteristic diagram of the diameter of the polyester thread through a polyester thread diameter parser based on a decoder to obtain a decoded result of the diameter data of the polyester thread; S142, calculating the cross-sectional area of the polyester thread based on the decoded result of the diameter data of the polyester thread.

[0071] Specifically, in step S141, the polyester thread diameter state feature map is passed through a polyester thread diameter parser based on a decoder to obtain a decoded result of polyester thread diameter data. It should be understood that although the enhanced polyester thread diameter state feature map after the dynamic receptive field enhancement already contains multi-scale thread diameter morphological information (such as local curvature, axial gradient, texture anisotropy), its essence is still a high-dimensional abstract feature vector and cannot be directly used for mechanical model calculations. Therefore, in the technical solution of this application, the enhanced polyester thread diameter state feature map is further passed through a polyester thread diameter parser based on a decoder to obtain a decoded result of polyester thread diameter data. The function of the decoder is to establish a spatial mapping relationship from pixel-level features to physical thread diameter values, thereby generating a corresponding decoded result of polyester thread diameter data. Specifically, the decoder first performs feature extraction and feature fusion on the input polyester thread diameter state feature map through convolution operations. These convolution operations are aimed at further extracting key information in the polyester thread diameter state feature map and fusing feature information at different levels. The convolution operation slides a convolution kernel on the polyester thread diameter state feature map and performs weighted summation on local features to extract higher-level features. These convolution operations can not only capture the spatial relationship between local features but also enhance the representation ability of features. Next, the decoder gradually enlarges the resolution of the polyester thread diameter state feature map through upsampling operations. Upsampling operations usually use transposed convolution operations or interpolation methods, such as nearest neighbor interpolation, bilinear interpolation, etc. These upsampling methods can enlarge the low-resolution polyester thread diameter state feature map to a higher resolution while retaining the key information in the polyester thread diameter state feature map. During the enlargement process, the decoder converts the features in the polyester thread diameter state feature map into specific thread diameter values through the learned mapping relationship. This mapping relationship is learned through a large amount of training data, enabling the decoder to accurately decode the information in the polyester thread diameter state feature map into thread diameter data. During the upsampling process, the decoder also needs to introduce skip connections to enhance the representation ability of features. Skip connections are a technique for fusing features in the decoder with features in the encoder. Through skip connections, the decoder can directly utilize the low-level feature information extracted in the encoder, thereby enhancing the representation ability of features. These low-level feature information contains more detailed information and can help the decoder more accurately recover the thread diameter data of the polyester thread. After a series of convolution operations and upsampling operations, the features obtained by the decoder already contain relatively accurate thread diameter information. However, this thread diameter information still exists in the form of feature vectors and needs to be further converted into specific thread diameter values. This conversion process is usually achieved through a fully connected layer or a convolutional layer. The role of the fully connected layer or the convolutional layer is to further compress and convert the information in the feature vector and finally output specific thread diameter values.This process can be achieved through the learned weight and bias parameters, enabling the decoder to accurately convert the information in the feature vector into the decoding result of the polyester thread diameter data. Among them, the decoding result of the polyester thread diameter data specifically represents the quantitative information of the actual diameter size at different axial positions of the polyester thread during the tensile test. This data is presented in a continuous numerical form, reflecting the diameter size of the polyester thread at each corresponding pixel position in the image, thus being able to precisely depict the diameter distribution of the polyester thread during the tensile process. Through this decoding result, the thickness changes of the polyester thread at different positions can be clearly observed, providing a key physical quantity basis for subsequent tensile strength calculation and performance evaluation.

[0072] Specifically, in step S142, calculate the cross-sectional area of the polyester thread based on the decoding result of the polyester thread diameter data. It should be understood that the tensile strength of the polyester thread is calculated by the ratio of the applied tensile force to the cross-sectional area of the polyester thread, and a slight change in the thread diameter will cause a significant change in the cross-sectional area, thereby affecting the calculation result of the tensile strength. Considering that the tensile strength is not only related to the applied tensile force but also closely related to the cross-sectional area of the polyester thread. Therefore, it is necessary to calculate the cross-sectional area of the polyester thread based on the decoding result of the polyester thread diameter data. Among them, when calculating the cross-sectional area based on the decoding result, the axial integration algorithm is used to perform calculus operations on the continuously distributed diameter data, thereby generating the corresponding decoding result of the polyester thread diameter data and solving the problem of strength evaluation deviation caused by distorted cross-sectional area calculation.

[0073] In the above intelligent detection method for the tensile strength of polyester thread, in step S150, the tensile force value data of the polyester thread sample to be detected is collected through a tensile force sensor. It should be understood that the tensile strength is a key index to measure the material properties of polyester thread, and its calculation depends on two core parameters: the tensile force applied to the polyester thread and the cross-sectional area of the polyester thread. As a key device for measuring the tensile force, the tensile force sensor is used to provide accurate and reliable tensile force values, thus providing a direct and accurate physical quantity basis for the calculation of the tensile strength. The definition of tensile strength is the ratio of the tensile force applied to the material to the cross-sectional area of the material. This definition clearly points out the importance of the tensile force in the calculation of the tensile strength. During the actual detection process, the polyester thread sample is subjected to a gradually increasing tensile force in the tensile testing machine until it breaks. The role of the tensile force sensor is to monitor and record in real time the tensile force values applied to the polyester thread during this process. These tensile force value data are continuous and dynamic, and can reflect the actual force magnitude borne by the polyester thread at different stretching stages. The working principle of the tensile force sensor is based on the mechanical properties of materials. When a tensile force acts on the sensor, the elastic element inside the sensor will deform. This deformation is proportional to the applied tensile force. Through strain gauges or other sensitive elements, the deformation can be converted into an electrical signal. These electrical signals are processed through amplification, filtering, and analog-to-digital conversion, and finally accurate tensile force values are obtained. The accuracy and sensitivity of the tensile force sensor directly affect the accuracy of the tensile force measurement, and thus affect the reliability of the tensile strength calculation. Considering that the traditional method for detecting the tensile strength usually relies on manual reading of the tensile force value, this method is not only inefficient but also easily affected by human factors, resulting in large measurement errors. The use of a tensile force sensor can achieve automatic and accurate measurement of the tensile force value. Through a high-precision tensile force sensor, the tensile force change of the polyester thread during the stretching process can be recorded in real time and continuously, thus providing accurate data support for the calculation of the tensile strength. Moreover, the tensile force sensor can provide dynamic tensile force data. During the stretching process of the polyester thread, the tensile force gradually increases, and the tensile force sensor can monitor this change process in real time and record the tensile force value at a high frequency. This dynamic monitoring ability makes the calculation of the tensile strength more accurate because it can reflect the actual force-bearing situation of the polyester thread at different stretching stages. Especially when approaching the breaking point of the polyester thread, the change of the tensile force value is more critical, and the tensile force sensor can accurately capture the tensile force value at this moment, thus providing accurate data for calculating the maximum tensile strength of the polyester thread. In addition, considering that in the tensile test, the polyester thread may break suddenly, generating a large impact force. The tensile force sensor can withstand this impact force and can monitor the change of the tensile force in real time, thus avoiding equipment damage or personal injury caused by excessive tensile force. At the same time, the high-precision measurement ability of the tensile force sensor can also reduce misjudgment caused by measurement errors and improve the credibility of the detection results.Specifically, in a specific embodiment of the present application, the installation position of the tensile sensor should be as close as possible to the polyester thread sample to reduce the moment error caused by too long a force arm. And before using the tensile sensor, calibration is required to ensure the accuracy of the measurement results of the tensile sensor. The calibration process usually includes zero calibration and range calibration of the sensor. Zero calibration is to adjust the output of the sensor so that it outputs zero under no load. Range calibration is to apply a known force value and adjust the output of the sensor so that it corresponds to the applied force value. It is worth mentioning that during the stretching process of the polyester thread, the tensile sensor converts the real-time measured force value into an electrical signal, which is collected and processed by the data acquisition system. The data acquisition system usually includes a signal amplifier, a filter, and an analog-to-digital converter (ADC). The function of the signal amplifier is to amplify the weak electrical signal output by the sensor to a range suitable for analog-to-digital conversion. The filter is used to remove noise and interference in the signal and improve the signal quality. The analog-to-digital converter converts the amplified and filtered analog signal into a digital signal for further processing and analysis by a computer or other data processing devices. The sampling frequency of the data acquisition system should be high enough to ensure that the rapid changes in the force value during the stretching process of the polyester thread can be captured, especially the force value changes near the breaking point.

[0074] In the above intelligent detection method for the tensile strength of polyester thread, in step S160, the tensile force value data is divided by the cross-sectional area of the polyester thread to obtain the tensile strength data of the polyester thread sample object to be detected. It should be understood that the tensile strength reflects the maximum stress that the material can withstand when subjected to a tensile force. Stress is the force per unit area, which is defined as the ratio of the acting force to the force-bearing area. In the tensile test of polyester thread, the tensile force value data is directly measured by a tensile force sensor, while the cross-sectional area of the polyester thread is calculated from the wire diameter data. Dividing the tensile force value by the cross-sectional area of the polyester thread, the result obtained is the tensile strength. This result can eliminate the strength calculation error caused by the different thicknesses of polyester threads, thereby providing an index directly related to the actual performance of the material. In the actual detection process, the wire diameter of the polyester thread is not completely uniform but has certain variations. This non-uniformity of the wire diameter will result in different cross-sectional areas at different positions of the polyester thread, which in turn affects the stress it bears during the tensile process. If only the tensile force value is used as the strength index while ignoring the change in wire diameter, the calculated strength value will not be able to truly reflect the actual strength characteristics of the polyester thread. Therefore, it is necessary to calculate the cross-sectional area of the polyester thread and divide the tensile force value by this cross-sectional area to obtain accurate tensile strength data. It is worth mentioning that the tensile strength is directly proportional to the tensile force value and inversely proportional to the cross-sectional area. Therefore, even under the same tensile force, polyester thread samples with different cross-sectional areas will exhibit different tensile strengths. By calculating the cross-sectional area and using it for strength calculation, it can ensure the comparability of the strength data between different samples, thus providing an objective and accurate basis for the quality assessment of polyester thread. The calculation result of the tensile strength is of great significance for the quality control and performance evaluation of polyester thread. By comparing the tensile strength data of polyester thread samples from different batches or different production processes, it can be determined whether their quality meets the standard requirements. For example, if the tensile strength of a certain batch of polyester thread samples is significantly lower than the standard value, it can be inferred that there may be quality problems in this batch of polyester thread, such as impure raw materials, production process defects, etc. In addition, the tensile strength data can also be used to optimize the production process of polyester thread. By adjusting the production process parameters, the tensile strength of the polyester thread can be increased, thereby improving its product quality and performance. It is worth mentioning that since this application synchronously collects the sequential force value data and dynamic wire diameter data during the tensile process in real time, the tensile strength data here can refer to the real-time tensile strength during the tensile process.

[0075] In summary, the intelligent detection method for the tensile strength of polyester threads based on the embodiments of the present application is elucidated. It first captures the multi-scale morphological characteristics of the polyester threads in real time during the stretching process, and enhances the local detail representation of the non-uniform regions of the polyester thread diameters under the regulation of dynamic receptive fields. Then, through the feature decoder, the pixel-level spatial state feature information of the polyester thread diameters is mapped into continuous distribution data of the diameters, synchronously correlates the time-series force value signals of the tensile force sensors, establishes a cross-modal mapping relationship of "image - diameter - force value", and finally realizes the adaptive compensation calculation of the tensile strength based on the cross-sectional area dynamic correction model to eliminate the coupled interference of diameter fluctuations and tension discreteness on strength evaluation, break through the dependence of traditional detection on the assumption of material uniformity, and provide a basis for the detection and performance evaluation of the tensile strength of polyester threads.

[0076] Figure 5 FIG. is a system block diagram of an intelligent detection system for the tensile strength of polyester threads according to an embodiment of the present application. As Figure 5 shown, the intelligent detection system 100 for the tensile strength of polyester threads according to an embodiment of the present application includes: a stretching detection image acquisition module 110 for polyester thread sample objects, configured to acquire stretching detection images of the polyester thread sample objects to be detected through a high-resolution camera; a polyester thread target region of interest recognition and feature extraction module 120, configured to perform recognition and feature extraction of the polyester thread target region of interest on the stretching detection images to obtain polyester thread diameter state features; a polyester thread diameter state feature significant enhancement processing module 130, configured to perform significant enhancement processing of the polyester thread diameter state features based on dynamic receptive fields to obtain enhanced polyester thread diameter state features; a polyester thread cross-sectional area calculation module 140, configured to determine polyester thread diameter data based on the enhanced polyester thread diameter state features, and calculate the cross-sectional area of the polyester threads according to the polyester thread diameter data; a stretching force value data acquisition module 150 for the polyester thread sample objects to be detected, configured to acquire stretching force value data of the polyester thread sample objects to be detected through a tensile force sensor; and a stretching strength data acquisition module 160 for the polyester thread sample objects to be detected, configured to divide the stretching force value data by the cross-sectional area of the polyester threads to obtain the stretching strength data of the polyester thread sample objects to be detected. Here, those skilled in the art can understand that the specific operations of each step in the above intelligent detection system for the tensile strength of polyester threads have been described in detail above with reference to Figures 1 to 4 the description of the intelligent detection method for the tensile strength of polyester threads, and therefore, the repeated description thereof will be omitted.

[0077] As described above, the intelligent detection system 100 for the tensile strength of polyester yarn according to the embodiments of the present application can be implemented in various terminal devices. In one example, the intelligent detection system 100 for the tensile strength of polyester yarn can be integrated into the terminal device as a software module and / or a hardware module. For example, the intelligent detection system 100 for the tensile strength of polyester yarn can be a software module in the operating system of the terminal device, or can be an application program developed for the terminal device; of course, the intelligent detection system 100 for the tensile strength of polyester yarn can also be one of the many hardware modules of the terminal device.

[0078] Alternatively, in another example, the intelligent detection system 100 for the tensile strength of polyester yarn and the terminal device can also be separate devices, and the intelligent detection system 100 for the tensile strength of polyester yarn can be connected to the terminal device through a wired and / or wireless network, and transmit and interact information according to a predefined data format.

[0079] In summary, the intelligent detection system for the tensile strength of polyester yarn based on the embodiments of the present application is elucidated. It first captures the multi-scale morphological characteristics of the polyester yarn during the stretching process in real time, and enhances the local detail representation of the non-uniform area of the polyester yarn diameter under the regulation of the dynamic receptive field. Then, through the feature decoder, the pixel-level polyester yarn diameter spatial state feature information is mapped into the continuous distribution data of the diameter, and the timing force value signal of the tensile force sensor is synchronously correlated to establish a cross-modal mapping relationship of "image - diameter - force value". Finally, based on the cross-sectional area dynamic correction model, the adaptive compensation calculation of the tensile strength is realized to eliminate the coupling interference of the diameter fluctuation and the tension discreteness on the strength evaluation, break through the dependence of the traditional detection on the assumption of material uniformity, and provide a basis for the detection and performance evaluation of the tensile strength of polyester yarn.

Claims

1. An intelligent detection method for the tensile strength of polyester thread, characterized in that, Including: Collecting tensile test images of the detected polyester thread sample object through a high-resolution camera; Identifying the region of interest of the polyester thread target and extracting features from the tensile test image to obtain the polyester thread diameter state features; Performing significant enhancement processing on the polyester thread diameter state features based on a dynamic receptive field to obtain enhanced polyester thread diameter state features; Based on the enhanced polyester thread diameter state features, determining the polyester thread diameter data and calculating the cross-sectional area of the polyester thread according to the polyester thread diameter data; Collecting tensile force value data of the detected polyester thread sample object through a tensile force sensor; Dividing the tensile force value data by the cross-sectional area of the polyester thread to obtain the tensile strength data of the detected polyester thread sample object.

2. The intelligent detection method for the tensile strength of polyester yarn according to claim 1, wherein Identifying the region of interest of the polyester thread target and extracting features from the tensile test image to obtain the polyester thread diameter state features, including: Performing image denoising and grayscale processing on the tensile test image to obtain a grayscale tensile test image; Extracting the region of interest of the polyester thread target from the grayscale tensile test image to obtain a region of interest image of the polyester thread target; Extracting the polyester thread diameter state features from the region of interest image of the polyester thread target to obtain the polyester thread diameter state features.

3. The intelligent detection method for the tensile strength of polyester thread according to claim 2, characterized in that, Extracting the polyester thread diameter state features from the region of interest image of the polyester thread target to obtain the polyester thread diameter state features, including: passing the region of interest image of the polyester thread target through a polyester thread diameter state feature extractor based on DenseNet to obtain a polyester thread diameter state feature map as the polyester thread diameter state features.

4. The intelligent detection method for the tensile strength of polyester thread according to claim 3, characterized in that, Performing significant enhancement processing on the polyester thread diameter state features based on a dynamic receptive field to obtain enhanced polyester thread diameter state features, including: Performing information compression based on feature decoupling on the polyester thread diameter state feature map to obtain a polyester thread diameter state feature vector to be enhanced and distilled; Based on the spatial distribution structure of the polyester thread diameter state feature vector to be enhanced and distilled, screening out a set of pixel-level polyester thread diameter state feature vectors within the local receptive field from the set of pixel-level polyester thread diameter state feature vectors; Based on the set of pixel-level polyester thread diameter state feature vectors within the local receptive field, performing significance enhancement on the polyester thread diameter state feature vector to be enhanced to obtain an enhanced polyester thread diameter state feature vector; Aggregating multiple enhanced polyester thread diameter state feature vectors to obtain an enhanced polyester thread diameter state feature map as the enhanced polyester thread diameter state features.

5. The intelligent detection method for the tensile strength of polyester yarn according to claim 4, characterized in that Performing information compression based on feature decoupling on the polyester thread diameter state feature map to obtain a polyester thread diameter state feature vector to be enhanced and distilled, including: Performing feature decoupling on the polyester thread diameter state feature map along the channel dimension to obtain a set of pixel-level polyester thread diameter state feature vectors; Extracting the pixel-level polyester thread diameter state feature vector at the (i, j) pixel position from the set of pixel-level polyester thread diameter state feature vectors as the polyester thread diameter state feature vector to be enhanced. Perform information compression on the line diameter state feature vector of the polyester thread to be strengthened to obtain the line diameter state feature vector of the polyester thread to be strengthened and distilled.

6. The intelligent detection method for the tensile strength of polyester yarn according to claim 5, wherein, Based on the spatial distribution structure of the line diameter state feature vector of the polyester thread to be strengthened and distilled, select the set of pixel-level polyester thread line diameter state feature vectors within the local receptive field from the set of pixel-level polyester thread line diameter state feature vectors, including: Based on the characteristic distribution spatial structure characteristics of the line diameter state feature vector of the polyester thread to be strengthened and distilled, determine the size of the characteristic receptive field of the line diameter state feature vector of the polyester thread to be strengthened. Based on the size of the characteristic receptive field, select the set of pixel-level polyester thread line diameter state feature vectors within the local receptive field from the set of pixel-level polyester thread line diameter state feature vectors.

7. The intelligent detection method for the tensile strength of polyester thread according to claim 6, characterized in that, The strengthened polyester thread line diameter state feature vector is the channel feature vector at the (i, j) pixel position of the strengthened polyester thread line diameter state feature map.

8. The intelligent detection method for the tensile strength of polyester yarn according to claim 7, characterized in that Based on the set of pixel-level polyester thread line diameter state feature vectors within the local receptive field, perform saliency enhancement on the line diameter state feature vector of the polyester thread to be strengthened to obtain the strengthened polyester thread line diameter state feature vector, including: Perform conformal representation fusion on the set of pixel-level polyester thread line diameter state feature vectors within the local receptive field to obtain the pixel-level polyester thread line diameter state fusion feature vector within the local receptive field. Perform weighted fusion on the pixel-level polyester thread line diameter state fusion feature vector within the local receptive field and the line diameter state feature vector of the polyester thread to be strengthened to obtain the strengthened polyester thread line diameter state feature vector.

9. The intelligent detection method for the tensile strength of polyester thread according to claim 8, characterized in that, Based on the strengthened polyester thread line diameter state feature, determine the polyester thread line diameter data, and calculate the cross-sectional area of the polyester thread according to the polyester thread line diameter data, including: Pass the polyester thread line diameter state feature map through a polyester thread line diameter parser based on a decoder to obtain the decoding result of the polyester thread line diameter data. Calculate the cross-sectional area of the polyester thread based on the decoding result of the polyester thread line diameter data.

10. An intelligent detection system for the tensile strength of polyester threads, characterized in that, Including: A stretching detection image acquisition module for the polyester thread sample object, which is used to acquire the stretching detection image of the polyester thread sample object to be detected through a high-resolution camera. A polyester thread target region of interest recognition and feature extraction module, which is used to perform polyester thread target region of interest recognition and feature extraction on the stretching detection image to obtain the polyester thread line diameter state feature. A polyester thread line diameter state feature saliency enhancement processing module, which is used to perform polyester thread line diameter state feature saliency enhancement processing based on a dynamic receptive field on the polyester thread line diameter state feature to obtain the strengthened polyester thread line diameter state feature. A polyester thread cross-sectional area calculation module, which is used to determine the polyester thread line diameter data based on the strengthened polyester thread line diameter state feature, and calculate the cross-sectional area of the polyester thread according to the polyester thread line diameter data. A stretching force value data acquisition module for the polyester thread sample object to be detected, which is used to acquire the stretching force value data of the polyester thread sample object to be detected through a tensile sensor. A stretching strength data acquisition module for the polyester thread sample object to be detected, which is used to divide the stretching force value data by the cross-sectional area of the polyester thread to obtain the stretching strength data of the polyester thread sample object to be detected.

Images