Cloth spinning position correction method and system based on image visual identification

Abstract

The invention relates to the field of image recognition, in particular to a cloth spinning position correction method and system based on image visual recognition. The method comprises the following steps: acquiring a cloth real-time monitoring image on a spinning device based on a high-resolution industrial camera; performing global image brightness enhancement and end-to-end super-resolution reconstruction on the cloth real-time monitoring image to construct a super-resolution reconstructed cloth image; performing key cloth edge visual identification on the super-resolution reconstructed cloth image, and performing edge point dynamic change tracking so as to obtain a cloth edge point dynamic movement track; carrying out edge point accurate spatial position calculation on the cloth edge point dynamic movement track, and carrying out cloth position deviation calculation so as to generate cloth position deviation trend data; and obtaining the material parameters of the current textile cloth. According to the invention, highly automatic and accurate cloth position deviation correction is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a method and system for correcting cloth weaving position based on image visual recognition. Background Art

[0002] With the continuous development of the textile industry and the continuous improvement of automation level, cloth weaving devices have been widely used in various production lines. Especially in large-scale production, the use of automated weaving devices has greatly improved production efficiency, but with it comes the increasingly prominent problem of cloth position deviation. The deviation of cloth position is caused by many factors, such as equipment wear, uneven fabric material, changes in tensile force, and differences in cloth thickness. These factors will affect the precise alignment of cloth in the weaving process, thereby affecting product quality and even leading to production stagnation or equipment failure.

[0003] Traditional cloth position correction methods mainly rely on mechanical sensors, physical probes or manual visual inspection to monitor and correct cloth deviations in real time. Although these methods can detect cloth position deviations to a certain extent, they have problems such as slow response speed, low accuracy, and cumbersome operation. Especially in the high-speed transmission process of cloth, traditional correction methods often cannot meet the requirements of high efficiency and high accuracy, and are easily disturbed by environmental factors, resulting in error accumulation and frequent equipment downtime. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a method and system for correcting cloth weaving position based on image visual recognition, so as to solve at least one of the above technical problems.

[0005] To achieve the above object, the present invention provides a method for correcting cloth weaving position based on image visual recognition, comprising the following steps: Step S1: acquiring a real-time monitoring image of cloth on a textile device based on a high-resolution industrial camera; performing global image brightness enhancement and end-to-end super-resolution reconstruction on the real-time monitoring image of the cloth to construct a super-resolution reconstructed cloth image; Step S2: performing visual recognition of key cloth edges on the super-resolution reconstructed cloth image, and tracking the dynamic changes of edge points, thereby obtaining the dynamic movement trajectory of the cloth edge points; Step S3: Calculate the precise spatial position of the edge points based on the dynamic movement trajectory of the edge points of the cloth, and calculate the cloth position deviation, so as to generate cloth position deviation trend data; Step S4: obtaining material parameters of the current textile cloth; performing real-time textile working condition mining on the material parameters of the current textile cloth, and performing adaptive position deviation compensation calculation according to cloth position deviation trend data, thereby generating a cloth position deviation compensation value; Step S5: performing cloth deviation correction simulation on the textile device according to the cloth position deviation compensation value, and then performing secondary deviation accuracy correction, thereby generating cloth position deviation accuracy optimization parameters; Step S6: Perform instant deviation correction delay analysis based on the cloth position deviation accuracy optimization parameters, and perform position correction iterative learning to build an intelligent cloth deviation correction control model.

[0006] The present invention not only improves the visibility of the image, but also removes noise and improves the effective information of the image by performing global image brightness enhancement on low-light or blurred areas. The super-resolution reconstruction technology improves the detail resolution of the image through a deep learning algorithm, so that the tiny displacement, stretching and deformation of the cloth surface can be accurately identified, reducing the deviation error caused by low image resolution. The end-to-end super-resolution reconstruction is directly processed at the image acquisition end, reducing the delay of data transmission and calculation, so that the cloth position correction process can be carried out in real time. The edge vision recognition algorithm can accurately identify the edge of the cloth, especially in the case of complex textures and diversified fabrics, and can accurately find the edge information of the cloth. This provides stable and reliable preliminary data for subsequent dynamic tracking. By dynamically tracking the edge points, the position changes of the cloth in the weaving process are monitored in real time, and the actual movement trajectory of the cloth is tracked. This process is particularly important because the displacement of the cloth in the weaving process is usually accompanied by complex deformation and physical stretching, and dynamic tracking can provide more accurate cloth movement information. By accurately calculating the spatial position of the edge points, the real-time spatial coordinates of the cloth on the textile equipment are accurately obtained, avoiding the position measurement error caused by the deformation or displacement of the cloth. Based on the movement trajectory of the edge points of the cloth, combined with time series analysis, the dynamic deviation of the cloth is trend predicted and calculated in real time. This can help the system identify potential production problems in advance, such as cloth deviation trends or position error accumulation problems, so that timely adjustments can be made before the problems expand. Obtaining the material parameters of the cloth (such as thickness, elasticity, density, etc.) is crucial for accurately controlling the movement and position of the cloth. These material properties will affect the stretching, deformation and response speed of the cloth, so the real-time acquisition and analysis of these parameters can provide strong data support for position adjustment. Real-time mining and analysis of the physical changes and movement laws of the cloth under the current textile working conditions, combined with the position deviation trend data, more accurately calculate the deformation degree and position correction requirements of the cloth in a specific environment. Through the simulation calculation of the compensation value of the cloth position deviation, virtual verification is carried out before the actual correction to ensure the feasibility of the adjustment strategy. Through the correction operation in the simulated environment, the effects of different compensation schemes are predicted, thereby reducing the risks in actual operation. After the simulation adjustment, a secondary deviation accuracy correction is performed to ensure that the position deviation of all cloth can be fine-tuned and optimized. By analyzing the delay of real-time correction, identifying and optimizing the time lag problem in the control system, we can ensure that the cloth position adjustment can respond immediately, thereby minimizing the loss of production efficiency caused by delays. Through iterative learning of position correction, the system can continuously accumulate data and optimize the adjustment strategy in practice, making each correction operation more accurate and efficient. This self-learning mechanism brings continuous performance improvement to the system.Through iterative learning and data analysis, an intelligent cloth deviation correction control model was eventually established, enabling the system to automatically adjust control parameters according to different cloth types, production conditions and working requirements, thereby achieving highly automated and precise cloth position correction.

[0007] Preferably, step S1 comprises the following steps: Step S11: obtaining a real-time monitoring image of cloth on a textile device based on a high-resolution industrial camera; Step S12: performing global image brightness enhancement on the real-time cloth monitoring image to obtain an image brightness enhanced cloth image; Step S13: performing edge texture detail recognition on the cloth image with enhanced image brightness to extract the edge texture details of the cloth; Step S14: performing edge texture sharpening based on the edge texture details of the cloth, thereby generating an edge detail optimized cloth image; Step S15: performing end-to-end super-resolution reconstruction on the edge detail optimized cloth image to construct a super-resolution reconstructed cloth image.

[0008] The present invention makes the details of the cloth more prominent by brightness enhancement, especially in the darker or blurred areas in the image, increases the sensitivity to the subtle changes of the cloth surface, and is helpful for subsequent edge detection and texture analysis. In the cloth weaving process, the image acquisition environment has uneven light or is too dark. Brightness enhancement can improve the image under these low-light conditions and make the image content clearer. This step is particularly effective for monitoring at night or in dim environments. Through the edge texture detail recognition algorithm, the system can accurately extract the edge features and texture details of the cloth, which is crucial for judging the actual position and shape of the cloth. Especially when the deformation or displacement of the cloth is small and the traditional image processing method cannot effectively recognize it, the extracted edge texture is sharpened to further highlight the key details of the cloth surface and enhance the resolution and clarity of the image, which is crucial for accurately identifying the edge and slight position changes of the cloth. Especially when the contrast between the edge and the background is low, super-resolution reconstruction can restore the lost details in the image, especially the restoration and refinement of the cloth surface texture, so that the slight deviation, damage or deformation in the image is more clearly displayed, which is convenient for accurate correction calculation.

[0009] Preferably, step S15 specifically comprises the following steps: Perform multi-scale convolution decomposition on the edge detail optimized cloth image to extract image convolution features of different scales; Perform principal component analysis on image convolution features of different scales to extract key convolution feature vectors of the image; A convolutional neural network architecture is used, wherein the convolutional neural network architecture includes multiple convolutional layers, pooling layers, and activation layers; The key convolution feature vector of the image is input into the convolutional neural network architecture for end-to-end deep learning to generate end-to-end convolutional visual features; Back-propagation optimization is performed on the end-to-end convolutional visual features to generate parameters that minimize the image convolution loss; Perform super-resolution iterative training on the parameters that minimize the image convolution loss to obtain image super-resolution iterative optimization parameters; The edge detail optimized cloth image is subjected to full-image super-resolution reconstruction according to the image super-resolution iterative optimization parameters to construct a super-resolution reconstructed cloth image.

[0010] The present invention extracts features of different scales from images through multi-scale convolution decomposition, which is crucial for processing different textures, edges and details on cloth. The surface of textile cloth contains textures of different sizes and directions. The use of multi-scale convolution can more comprehensively capture different levels of information of cloth, such as subtle wrinkles and local deformation of textiles. After the convolution features at multiple scales are processed by PCA, the key feature vectors obtained will reduce the redundancy of input data, making subsequent deep learning processing more efficient, which not only saves computing resources but also accelerates the training process of the model. The combination of multi-layer convolution layers and pooling layers can gradually extract different levels of information of the image. The pooling layer reduces the amount of data when extracting features, helps remove unnecessary information, and retains the main features of the image, so that the network can focus on key information. By directly inputting the extracted key convolution feature vectors into the CNN architecture for end-to-end deep learning, the network automatically optimizes all parameters during the training process to achieve end-to-end processing of image recognition. This makes the process from feature extraction to final output seamless, avoiding the complexity of manual feature selection and intermediate steps. By optimizing the convolution loss function, the system can accurately adjust the position deviation of the cloth in the image, gradually correct the error, and ensure that the subsequent image analysis and cloth position adjustment are more accurate. Back propagation optimization enables the network to not only adapt to the current data, but also have a certain generalization ability, and can cope with various situations such as different cloth types and environmental changes. Super-resolution training enhances the resolution of the image through an iterative process, making the details in the image clearer. This step is crucial for the accurate identification of slight changes and local deformations of the cloth, especially when the cloth position deviation is very subtle. Super-resolution effectively improves image quality, and uses optimized super-resolution iterative training parameters to reconstruct the entire image, significantly improving the resolution and details of the entire cloth image, making every part of the cloth clearer, whether it is the edge of the cloth, wrinkles or slight deformations, can be captured more accurately.

[0011] Preferably, step S2 specifically comprises the following steps: Step S21: performing key cloth edge visual recognition on the super-resolution reconstructed cloth image, and marking a plurality of key cloth edge points; Step S22: performing image time-series frame decomposition on the super-resolution reconstructed cloth image to extract the cloth image frame sequence; Step S23: performing cloth dynamic optical flow analysis frame by frame on the cloth image frame sequence to generate cloth dynamic optical flow change data; Step S24: Tracking the dynamic changes of the cloth edge points based on the dynamic optical flow change data of the cloth based on multiple key cloth edge points, so as to obtain the dynamic movement trajectory of the cloth edge points.

[0012] After super-resolution reconstruction, the clarity of the image is significantly improved, and details can be more easily extracted from the cloth image, including small-scale deformations such as the edges and creases of the cloth. By marking multiple key edge points, the system can provide more detailed and accurate information for subsequent dynamic tracking. Through time-series frame decomposition, continuous video frames or image sequences are extracted from the dynamic behavior of the cloth, and the movement of the cloth during the weaving process can be tracked. This time-series frame extraction makes the cloth position correction no longer dependent on a single image, but can consider the movement trajectory of the cloth. Optical flow analysis is a method based on pixel motion estimation between continuous image sequences. By analyzing the pixel flow of the cloth in the image frame by frame, information such as the movement direction, speed and deformation of the cloth is obtained. This analysis helps to capture the dynamic changes of the cloth, especially in the process of cloth weaving, due to factors such as wrinkles, stretching or friction of the cloth, optical flow analysis can reveal the subtle movement of the cloth. Through dynamic optical flow data based on multiple key edge points, the system can track the dynamic movement trajectory of each edge point. By accurately tracking edge points, the position changes of the cloth in the textile equipment are monitored in real time, so that the cloth can be quickly responded to when it is offset to avoid large dislocation of the cloth. By tracking the dynamic changes of multiple edge points, the system can obtain comprehensive position change information of the cloth. This comprehensive dynamic tracking ensures accurate positioning and fine-tuning of the cloth position, especially in complex cloth dynamics, and can correct slight deviations in time.

[0013] Preferably, step S3 specifically comprises the following steps: Step S31: segmenting the dynamic movement trajectory of the edge point of the cloth into multiple time point trajectories, thereby obtaining the movement trajectories of the edge point at multiple time points; Step S32: Calculate the precise spatial position of the edge points based on the movement trajectories of the edge points at multiple time points to obtain the spatial positioning coordinates of the edge points at each time point; Step S33: Calculating the cloth position deviation of the edge point spatial positioning coordinates at each time point based on the preset cloth weaving standard position, and generating the cloth position deviation value at each time point; Step S34: Perform position deviation change analysis on the cloth position deviation value at each time point, thereby generating cloth position deviation trend data.

[0014] The present invention accurately captures the position changes of cloth at different times through multi-time point trajectory segmentation. In high-speed textile or cloth dynamic environment, the position of the edge point of the cloth will change slightly with time. Multi-time point segmentation ensures accurate tracking of the movement of the cloth at each time point, and divides the dynamic trajectory of the cloth edge point into multiple time periods, which is conducive to more detailed and in-depth trajectory analysis. During these segmented times, the system can identify the movement mode of the cloth at different stages, and provide higher resolution data for subsequent deviation calculation. The moving trajectory of the edge point at multiple times is used to calculate the spatial positioning coordinates of each time point. By calculating the accurate position of the edge point, the system can clearly understand the specific position of the cloth and avoid deviations caused by position estimation errors. By comparing the edge point position with the preset standard position, the cloth position deviation at each time point can be accurately calculated. The deviation of the cloth in the weaving process usually affects the quality of the final product. Timely detection and calculation of deviations are the key to avoiding quality problems. By analyzing the changing trend of the deviation value at each time point, the system can predict the future deviation changes of the cloth and make corrective measures in advance. This trend data analysis is crucial for scenarios such as long-term operation and automated control of cloth. It can effectively avoid the impact of long-term accumulated deviations on the cloth position. By analyzing the changing trend of the cloth position deviation, it helps identify potential system problems. If the cloth deviation value continues to increase, it indicates that there are problems such as abnormal cloth tension or equipment failure. Timely adjustments can effectively reduce production accidents.

[0015] Preferably, the specific steps of step S4 are: Step S41: obtaining material parameters of the current textile cloth; Step S42: performing cloth thickness identification on the material parameters of the current textile cloth to generate cloth thickness parameters; Step S43: performing fabric material analysis on the material parameters of the current textile cloth to generate textile cloth material characteristics; Step S44: calculating the cloth feeding speed of the textile device according to the cloth image frame sequence, and extracting the cloth feeding speed; Step S45: conducting real-time textile working condition mining on the cloth thickness parameter, the textile cloth material characteristics and the cloth feeding speed, thereby generating the current textile cloth working condition characteristics; Step S46: performing adaptive position deviation compensation calculation according to the cloth position deviation trend data and the current textile cloth working condition characteristics, thereby generating a cloth position deviation compensation value.

[0016] The present invention uses cloth thickness identification to analyze the different tension and motion resistance problems encountered by textile equipment when processing cloth of different thicknesses. The cloth thickness parameters can help the system adjust the working parameters of the equipment, thereby optimizing the operation effect and avoiding position deviation caused by thickness changes. The cloth thickness has a direct impact on the cloth feeding speed, tension control, winding and other processes. By measuring and identifying the thickness in real time, the system adjusts the corresponding textile parameters in real time to ensure that the cloth is always operating under the best working conditions in each link. The material characteristics (such as cotton, linen, silk, etc.) determine the softness, elasticity, stretchability and other behaviors of the cloth in the textile process. Material analysis helps the system identify the main physical properties of the cloth, thereby making accurate deviation correction decisions. Cloth of different materials requires different process parameters, such as different textile machine tension, speed and other controls. Material feature identification provides the system with real-time cloth type identification, thereby performing optimized control on each cloth, and the cloth in the textile process The cloth passes through the loom at a certain speed. The change of cloth feeding speed has a great influence on the tension and positioning of the cloth. Through the analysis of the image frame sequence, the system can extract the cloth feeding speed in real time and adjust the dynamic position control of the cloth based on this speed. Through the comprehensive analysis of the thickness, material characteristics and cloth feeding speed of the cloth, the system can dig out the overall working condition characteristics of the cloth in the textile process. This multi-dimensional analysis helps to fully understand the performance of the cloth under different textile conditions and provide comprehensive data support for deviation correction decisions. By combining the cloth position deviation trend data and the current textile working condition characteristics, the system can calculate and dynamically adjust the deviation compensation value of the cloth in real time, which enables the system to quickly compensate when the cloth position deviates to ensure that the cloth is always in the predetermined position. Taking into account the dynamic behavior of the cloth under different working conditions, the compensation value generated based on the real-time working condition can compensate for the deviation more accurately, avoid the inaccurate compensation caused by a single factor, and ensure that the position correction in the textile process is more accurate.

[0017] Preferably, the specific steps of step S5 are: Step S51: performing cloth deviation correction simulation on the textile device according to the cloth position deviation compensation value, and collecting cloth deviation correction simulation data; Step S52: performing secondary cloth position deviation identification on the cloth deviation correction simulation data according to the preset cloth weaving standard position, and extracting the cloth position deviation time point; Step S53: locating the position deviation edge point based on the cloth position deviation time point; Step S54: performing secondary deviation accuracy correction on the position deviation edge points, thereby generating cloth position deviation accuracy optimization parameters.

[0018] The present invention performs deviation correction simulation on the textile device through the cloth position deviation compensation value, and can provide early prediction for the position adjustment of the cloth in actual production. The simulation can verify the effectiveness of the cloth deviation correction parameters and avoid the accumulation of deviations caused by errors in actual operation. The data collection in the simulation process helps the system to continuously feedback the cloth position adjustment effect, and the data can be used for subsequent deviation analysis and correction. Through repeated simulation, the accuracy of cloth deviation correction can be optimized, and a more reliable deviation correction model can be formed through data accumulation. By analyzing the cloth deviation correction simulation data according to the preset standard position, the specific time point when the cloth deviates from the standard position can be accurately identified. Through this method, the system can efficiently detect each position deviation of the cloth in the textile process to ensure a clear understanding of the time and position of the deviation. It is understood that in the actual textile process, the deviation of the cloth usually does not occur instantaneously, but evolves gradually. By extracting the deviation time point, the system can promptly identify the slight change in the cloth position, and then take corresponding measures to avoid the accumulation of deviations and affect the production quality. Locating the deviation edge point of the cloth is a key step in identifying the key position of the deviation. By analyzing the deviation time point, the system can more accurately determine the deviation edge point of the cloth, that is, the part where the cloth position changes most dramatically, which helps to make the most precise adjustments in the entire textile process of the cloth. By performing secondary precision correction on the deviation edge point, the system can dynamically adjust the optimization parameters according to actual production data and simulation feedback, which enables the system to adapt to changes in different cloths, working conditions, equipment conditions, etc., and achieve adaptive and efficient deviation correction.

[0019] Preferably, the specific steps of step S6 are: Step S61: performing real-time cloth deviation correction control based on cloth position deviation accuracy optimization parameters and collecting cloth deviation correction response data; Step S62: performing instant deviation correction delay analysis on the cloth deviation correction response data to generate instant deviation correction feedback delay data; Step S63: optimizing the correction delay control on the instant correction feedback delay data to obtain correction delay optimization data; Step S64: Perform position correction iterative learning on the deviation correction delay optimization data to build an intelligent cloth deviation correction control model.

[0020] The present invention can make dynamic adjustments according to the instant feedback in the cloth movement process through real-time deviation correction control. This instant response improves the timeliness and accuracy of deviation correction and reduces further deviation caused by lag or error. Real-time deviation correction control is performed according to the optimized cloth position deviation accuracy parameters, which can quickly respond to the position deviation of the cloth in the textile process. Through the real-time control system, the cloth position deviation can be effectively reduced and accurately compensated. In the cloth deviation correction process, there is a certain time lag in the response of the cloth to the deviation correction control. By performing delay analysis on the deviation correction response data collected in real time, the system can identify the specific time point when the delay occurs and provide delay data for subsequent optimization. After analyzing the instant deviation correction feedback delay data, the system can optimize the deviation correction delay control strategy in a targeted manner. By adjusting the control strategy, the system can reduce the delay. The system can not only optimize the control strategy, but also perform real-time delay compensation during the correction process. Real-time compensation effectively eliminates the negative effects of feedback delay on cloth position adjustment, reduces error accumulation, and ensures the continuous accuracy of cloth position. The correction delay optimization data will be input into the position correction iterative learning, and the correction process will be continuously optimized using deep learning and adaptive algorithms. The system can improve the accuracy and adaptability of cloth position correction by continuously learning historical data and real-time feedback. Through iterative learning, the system can generate an intelligent cloth correction control model, which adjusts the control strategy based on a large amount of historical data and real-time data, thereby making the correction control more accurate, intelligent and efficient. The intelligent model will automatically learn and adapt to the textile needs of different cloths and different production environments.

[0021] In this specification, a cloth weaving position correction system based on image visual recognition is provided, which is used to execute the cloth weaving position correction method based on image visual recognition as described above, including: A super-resolution reconstruction module is used to obtain a real-time monitoring image of cloth on a textile device based on a high-resolution industrial camera; the real-time monitoring image of the cloth is subjected to global image brightness enhancement and end-to-end super-resolution reconstruction to construct a super-resolution reconstructed cloth image; an edge point movement trajectory module is used to perform key cloth edge visual recognition on the super-resolution reconstructed cloth image and track the dynamic changes of edge points to obtain the dynamic movement trajectory of the cloth edge points; The position deviation calculation module is used to calculate the precise spatial position of the edge points of the cloth according to the dynamic movement trajectory of the edge points, and calculate the cloth position deviation, so as to generate cloth position deviation trend data; The deviation compensation module is used to obtain the material parameters of the current textile cloth; conduct real-time textile working condition mining on the material parameters of the current textile cloth, and perform adaptive position deviation compensation calculation based on the cloth position deviation trend data, thereby generating a cloth position deviation compensation value; The secondary deviation correction module is used to simulate the cloth deviation correction of the textile device according to the cloth position deviation compensation value, and then perform secondary deviation accuracy correction to generate cloth position deviation accuracy optimization parameters; The deviation correction control module is used to perform instant deviation correction delay analysis based on the cloth position deviation accuracy optimization parameters, and to perform iterative learning of position correction to build an intelligent cloth deviation correction control model. The present invention improves the resolution of the image through super-resolution reconstruction, making the details and edges of the cloth clearer and providing more accurate image information for subsequent processing. The high-resolution image provides more accurate data for subsequent edge recognition and position calculation, thereby improving the reliability and correction accuracy of the entire system. The movement of the cloth is accurately captured through dynamic tracking of edge points, providing key data for subsequent deviation calculation. The movement trajectory of the edge points can help the system locate the position of the cloth more accurately and detect any potential deviations in time. Through the precise calculation of the dynamic trajectory of the edge points, the trend of the cloth position deviation is monitored in real time, and the moment when the cloth deviates from the standard position is identified in advance. The generated deviation trend data will provide a detailed adjustment basis for the subsequent deviation compensation module, ensuring that the correction operation is more accurate and timely. By acquiring the working conditions and material parameters of the cloth in real time, the compensation strategy is adaptively adjusted, thereby It can adapt to the needs of cloth correction under different materials and different working conditions. Through real-time analysis of the dynamic changes of cloth, it can accurately calculate the compensation value, reduce the impact caused by material differences or working condition fluctuations, and improve the overall stability of the system. The secondary deviation correction can further fine-tune on the basis of the first correction to ensure that the accuracy of the cloth position meets the expected standard. The secondary precision correction can effectively avoid the accumulation of errors, ensure that the cloth position is always within a reasonable deviation range, and improve the overall production quality. Through delay analysis, it can evaluate and reduce the delay in the correction process and improve the real-time response of the system. Through iterative learning and adaptive adjustment, the intelligent control model can automatically optimize the correction strategy, improve the system's adaptability in different production environments, and reduce human intervention. After continuous iterative training, the correction control model can gradually improve the long-term accuracy of cloth position correction and ensure the long-term stability of the production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A schematic flow chart of the steps of a method for correcting cloth weaving position based on image visual recognition according to the present invention; Figure 2 Detailed implementation flow chart of step S1; Figure 3 Detailed implementation flow chart of step S2; Figure 4 Detailed implementation flow chart of step S3. DETAILED DESCRIPTION

[0023] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0024] The present application example provides a method and system for correcting the position of a piece of cloth textile based on image visual recognition. The execution subject of the method and system for correcting the position of a piece of cloth textile based on image visual recognition includes but is not limited to: mechanical equipment, data processing platform, cloud server node, network upload device, etc. equipped with the system can be regarded as the general computing node of the present application, and the data processing platform includes but is not limited to: at least one of an audio image management system, an information management system, and a cloud data management system.

[0025] See also Figures 1 to 4 The present invention provides a method for correcting the position of cloth weaving based on image visual recognition, and the method for correcting the position of cloth weaving based on image visual recognition comprises the following steps: Step S1: acquiring a real-time monitoring image of cloth on a textile device based on a high-resolution industrial camera; performing global image brightness enhancement and end-to-end super-resolution reconstruction on the real-time monitoring image of the cloth to construct a super-resolution reconstructed cloth image; Step S2: performing visual recognition of key cloth edges on the super-resolution reconstructed cloth image, and tracking the dynamic changes of edge points, thereby obtaining the dynamic movement trajectory of the cloth edge points; Step S3: Calculate the precise spatial position of the edge points based on the dynamic movement trajectory of the edge points of the cloth, and calculate the cloth position deviation, so as to generate cloth position deviation trend data; Step S4: obtaining material parameters of the current textile cloth; performing real-time textile working condition mining on the material parameters of the current textile cloth, and performing adaptive position deviation compensation calculation according to cloth position deviation trend data, thereby generating a cloth position deviation compensation value; Step S5: performing cloth deviation correction simulation on the textile device according to the cloth position deviation compensation value, and then performing secondary deviation accuracy correction, thereby generating cloth position deviation accuracy optimization parameters; Step S6: Perform instant deviation correction delay analysis based on the cloth position deviation accuracy optimization parameters, and perform position correction iterative learning to build an intelligent cloth deviation correction control model.

[0026] The present invention not only improves the visibility of the image, but also removes noise and improves the effective information of the image by performing global image brightness enhancement on low-light or blurred areas. The super-resolution reconstruction technology improves the detail resolution of the image through a deep learning algorithm, so that the tiny displacement, stretching and deformation of the cloth surface can be accurately identified, reducing the deviation error caused by low image resolution. The end-to-end super-resolution reconstruction is directly processed at the image acquisition end, reducing the delay of data transmission and calculation, so that the cloth position correction process can be carried out in real time. The edge vision recognition algorithm can accurately identify the edge of the cloth, especially in the case of complex textures and diversified fabrics, and can accurately find the edge information of the cloth. This provides stable and reliable preliminary data for subsequent dynamic tracking. By dynamically tracking the edge points, the position changes of the cloth in the weaving process are monitored in real time, and the actual movement trajectory of the cloth is tracked. This process is particularly important because the displacement of the cloth in the weaving process is usually accompanied by complex deformation and physical stretching, and dynamic tracking can provide more accurate cloth movement information. By accurately calculating the spatial position of the edge points, the real-time spatial coordinates of the cloth on the textile equipment are accurately obtained, avoiding the position measurement error caused by the deformation or displacement of the cloth. Based on the movement trajectory of the edge points of the cloth, combined with time series analysis, the dynamic deviation of the cloth is trend predicted and calculated in real time. This can help the system identify potential production problems in advance, such as cloth deviation trends or position error accumulation problems, so that timely adjustments can be made before the problems expand. Obtaining the material parameters of the cloth (such as thickness, elasticity, density, etc.) is crucial for accurately controlling the movement and position of the cloth. These material properties will affect the stretching, deformation and response speed of the cloth, so the real-time acquisition and analysis of these parameters can provide strong data support for position adjustment. Real-time mining and analysis of the physical changes and movement laws of the cloth under the current textile working conditions, combined with the position deviation trend data, more accurately calculate the deformation degree and position correction requirements of the cloth in a specific environment. Through the simulation calculation of the compensation value of the cloth position deviation, virtual verification is carried out before the actual correction to ensure the feasibility of the adjustment strategy. Through the correction operation in the simulated environment, the effects of different compensation schemes are predicted, thereby reducing the risks in actual operation. After the simulation adjustment, a secondary deviation accuracy correction is performed to ensure that the position deviation of all cloth can be fine-tuned and optimized. By analyzing the delay of real-time correction, identifying and optimizing the time lag problem in the control system, we can ensure that the cloth position adjustment can respond immediately, thereby minimizing the loss of production efficiency caused by delays. Through iterative learning of position correction, the system can continuously accumulate data and optimize the adjustment strategy in practice, making each correction operation more accurate and efficient. This self-learning mechanism brings continuous performance improvement to the system.Through iterative learning and data analysis, an intelligent cloth deviation correction control model was eventually established, enabling the system to automatically adjust control parameters according to different cloth types, production conditions and working requirements, thereby achieving highly automated and precise cloth position correction.

[0027] In the embodiment of the present invention, refer to Figure 1 , is a schematic flow chart of a method of the present invention. In this example, the steps of the method include: Step S1: acquiring a real-time monitoring image of cloth on a textile device based on a high-resolution industrial camera; performing global image brightness enhancement and end-to-end super-resolution reconstruction on the real-time monitoring image of the cloth to construct a super-resolution reconstructed cloth image; In this embodiment, a suitable high-resolution industrial camera (such as 12 million pixels and above) is selected and installed at a suitable position of the textile device to ensure that it can capture the entire surface of the cloth. The camera should have good optical performance to reduce image distortion. The camera parameters, including shutter speed, exposure time, and gain, are configured to ensure that clear images can be obtained under different lighting conditions. The exposure time can be set to 1 / 100 second to adapt to fast-moving cloth. The image stream is transmitted to the computer for processing in real time through the USB or Ethernet interface of the camera. The image acquisition frequency is set to 30 frames per second to ensure real-time monitoring of the cloth status. Image processing software (such as OpenCV or MATLAB) is used for real-time image capture, and the timestamp of each frame of the image is recorded to facilitate subsequent analysis. Global image brightness enhancement technology is applied to the captured real-time monitoring image. Adaptive histogram equalization (CLAHE) is selected. This method can effectively improve the contrast of the image while maintaining details. CLAHE parameters, such as contrast limit (CLIP LIMIT) and grid size (TILE GRID SIZE), are set to optimize the enhancement effect. Generally, CLIP LIMIT can be set to 2.0, TILE GRID SIZE to 8x8 to adapt to images with different brightness distributions, remove noise from the enhanced image, use Gaussian filtering or median filtering to smooth the image, the kernel size of the Gaussian filter can be set to 5x5 to effectively remove random noise in the image without affecting the image details, select a suitable super-resolution reconstruction algorithm, such as SRCNN (Super-Resolution Convolutional Neural Network) or GAN (Generative Adversarial Network), both of which can effectively improve image resolution and generate clearer images. If SRCNN is selected, a high-quality training dataset must be prepared, including low-resolution and high-resolution image pairs, in order to train the model. Commonly used training datasets can include classic BSD300 or DIV2K. Use deep learning frameworks such as TensorFlow or PyTorch for model training, and set hyperparameters such as learning rate, batch size, and number of iterations. In general, the learning rate can be set to 0.001, the batch size is 16, and the number of training iterations is 1000. During the training process, the validation set is used to evaluate the model performance to ensure that the model can effectively restore the details of the low-resolution image. During the verification, indicators such as PSNR (Peak Signal-to-Noise Ratio) and SSIM (Structural Similarity Index) can be calculated. The goal is to make PSNR reach more than 30dB and SSIM close to 1. The real-time monitoring image is input into the trained super-resolution model to generate a super-resolution reconstructed cloth image. By comparing the low-resolution input image and the super-resolution output image, the reconstruction effect is evaluated, the characteristics of the reconstructed image, including resolution, contrast, and detail retention, are recorded, and a reconstruction result report is generated. .

[0028] Step S2: performing visual recognition of key cloth edges on the super-resolution reconstructed cloth image, and tracking the dynamic changes of edge points, thereby obtaining the dynamic movement trajectory of the cloth edge points; In this embodiment, the super-resolution reconstructed cloth image is further preprocessed to improve the accuracy of edge recognition. First, Gaussian filtering is applied to denoise, and the kernel size is set to 5x5 to reduce random noise in the image and ensure that the edge features are more obvious. Then, the image is grayed to convert the color image into a gray image to reduce the computational complexity while maintaining the edge information. A suitable edge detection algorithm is selected, such as the Canny edge detection algorithm. The Canny algorithm is widely used due to its excellent edge detection performance and can effectively detect subtle edges in the image. A low threshold and a high threshold are set in the Canny algorithm. Usually, the low threshold can be set to 50 and the high threshold can be set to 150 to balance the missed edge detection and false edge. The processed gray image is input into the Canny edge detection algorithm to generate an edge map of the cloth image. The coordinate information of each edge point is recorded and stored in a data structure (such as a list or array) for subsequent analysis and visualization of the detection results. The edge map is displayed through tools such as Matplotlib to ensure the accuracy and completeness of edge recognition. The optical flow method (Optical Flow) is used. Optical flow is used to track the dynamic changes of edge points. The optical flow method can track the movement of objects in the image through the pixel movement between consecutive frames. The Lucas-Kanade optical flow algorithm is selected. This algorithm has high computational efficiency and is suitable for motion tracking in a small range. It can handle the dynamic changes of edge points in image sequences. Prepare an image sequence (i.e., super-resolution reconstructed cloth images of consecutive frames), set the time interval between each frame (such as 1 / 30 second) for dynamic tracking, and use the Lucas-Kanade algorithm to calculate the motion vector of each edge point for each pair of consecutive frames. In this process, the window size is set to 5x5 to more accurately calculate the movement in the local area and record the initial position of each edge point. The dynamic movement trajectory of the edge points is generated based on the position changing with time. The trajectory of each edge point is visualized as a line segment to show its motion path in the image sequence. The dynamic trajectory is visualized using tools such as Matplotlib or OpenCV for further analysis and display. The dynamic behavior of the cloth can be observed intuitively by superimposing the motion trajectory of the edge points on the original image. The tracked dynamic trajectory of the edge points is analyzed, and the length, speed, acceleration and other characteristics of the trajectory are calculated to evaluate the dynamic behavior of the cloth. These indicators are used to judge the state changes of the cloth in the production process. Combined with the type of cloth and the production process, the behavioral characteristics of different cloths in the dynamic process are analyzed to provide data support for subsequent quality control and optimization.

[0029] Step S3: Calculate the precise spatial position of the edge points based on the dynamic movement trajectory of the edge points of the cloth, and calculate the cloth position deviation, so as to generate cloth position deviation trend data; In this embodiment, the coordinate information of each edge point in the image sequence is extracted. The coordinates of each edge point are usually expressed as two-dimensional coordinates (x, y). The coordinates of the edge points in each frame of the image are stored in a data structure (such as a list or array) to ensure that they can be easily accessed in subsequent calculations. The image coordinates are converted to actual space coordinates (such as millimeters or centimeters), and the intrinsic and extrinsic parameters of the camera must be considered. The camera calibration data is used for coordinate conversion to ensure that the spatial position of the edge point represents the real physical position. Set the focal length, sensor size, and image resolution of the camera to calculate the actual space coordinates. Assume that the focal length of the camera is 8mm, the image resolution is 1920x1080, and Z= f⋅d / h, where f is the focal length, d is the distance from the object to the camera, and h is the height of the object in the image. Determine the ideal position (reference position) of the cloth, which is usually set by process standards or production requirements. Record the coordinates of the reference position (x 0 , y 0 ), for example, set to (500, 300). The reference position is determined by the feature points in the initial frame image or the reference marks of the production equipment. For each edge point, calculate the deviation between its spatial position and the reference position. The deviation is calculated using the Euclidean distance, and the change of the deviation of each edge point during the entire monitoring process is recorded to analyze the dynamic behavior of the cloth. The deviation value of each edge point is recorded in the form of a time series to form deviation trend data. Each timestamp corresponds to a deviation value, and the data structure is stored using a data frame (such as PandasDataFrame) for subsequent analysis. Set the recording time interval (such as recording once per frame) to ensure that the deviation data can reflect the changes in the cloth during the dynamic process. Use visualization tools (such as Matplotlib or Seaborn) to generate a deviation trend chart to display the deviation data over time. By drawing a line chart or a bar chart, the dynamic change trend of the cloth position can be intuitively displayed. Record key data points in the trend chart, such as the maximum deviation, minimum deviation, and average deviation, for further quality assessment and process analysis.

[0030] Step S4: obtaining material parameters of the current textile cloth; performing real-time textile working condition mining on the material parameters of the current textile cloth, and performing adaptive position deviation compensation calculation according to cloth position deviation trend data, thereby generating a cloth position deviation compensation value; In this embodiment, suitable sensors (such as a tensile testing machine, a friction tester, and a thickness gauge) are selected to measure the material parameters of the cloth. The main parameters include the tensile strength, elastic modulus, friction coefficient, and thickness of the cloth. Experimental conditions, such as temperature and humidity, are set to ensure the repeatability of the measurement results. Usually, room temperature (about 20°C) and relative humidity (about 50%) are selected as standard test conditions. The measurement results of each material parameter are recorded and stored in a database or data frame to ensure that subsequent analysis can be easily accessed. The measurement data should include a timestamp and experimental conditions for tracking and analysis. The tensile strength is 300N (±5N), the elastic modulus is 1500MPa (±50MPa), the friction coefficient is 0.4, and the thickness is 1.2mm (±0.1mm). Textile working condition data is collected in real time through sensors (such as tension sensors, speed sensors, and temperature and humidity sensors) to ensure that the frequency of data collection (such as 10 times per second) can reflect the dynamic changes in the production process in real time. A data acquisition system (such as LabVIEW or Python data acquisition library) is used to connect the sensors to the data acquisition system. Data integration, to achieve real-time monitoring, analyze the real-time collected working condition data, extract key features (such as tension change rate, speed fluctuation and temperature and humidity changes), use data mining technology (such as cluster analysis or principal component analysis) to identify the characteristic patterns under different working conditions, combine material parameters, analyze the impact of working conditions on cloth performance, and provide a basis for subsequent position deviation compensation. According to the cloth position deviation trend data, an adaptive position deviation compensation model is established. The model should consider the impact of material parameters on dynamic characteristics in order to perform accurate compensation calculations, select a suitable compensation algorithm, such as PID controller or fuzzy controller, to achieve dynamic compensation, set algorithm parameters (such as proportional, integral, differential coefficients) to optimize the compensation effect, calculate the current compensation value based on the real-time collected cloth position deviation data and material parameters, analyze the calculated compensation value, evaluate its effectiveness in the dynamic compensation process, calculate compensation effect indicators, such as the root mean square error (RMSE) and maximum deviation of the deviation after compensation, combine real-time working condition data and material parameters, analyze the adaptability of the compensation strategy, and confirm whether the compensation effect meets production requirements.

[0031] Step S5: performing cloth deviation correction simulation on the textile device according to the cloth position deviation compensation value, and then performing secondary deviation accuracy correction, thereby generating cloth position deviation accuracy optimization parameters; In this embodiment, a cloth deviation correction model is established according to the compensation value obtained in the previous step. The model should take into account the dynamic characteristics of the cloth, material parameters, and feedback mechanism of the production equipment. Select a suitable simulation tool (such as MATLAB / Simulink) and input the compensation value and cloth motion parameters into the model for dynamic simulation. Determine the initial conditions of the simulation, including the initial position, velocity, and acceleration of the cloth. Set the initial position to (500, 300), the velocity to 1 m / s, and the simulation time to 5 seconds. Set the control parameters in the simulation, such as control gain and response time, to ensure the stability and accuracy of the deviation correction process. The control gain can be set to 1.5, and the response time is set to 0.5 seconds. Start the simulation program and observe the dynamic response of the cloth after applying the compensation value. Record the changes in the cloth position and the real-time feedback of the deviation value during the simulation. Display the simulation results through visualization tools and generate a dynamic trajectory diagram to facilitate the analysis of the deviation correction effect. Ensure that the simulation results can reflect the dynamic behavior in actual production. Based on the results of the deviation correction simulation, design a secondary deviation accuracy correction mechanism. This mechanism should be able to adjust the settings of the production equipment in real time to achieve higher accuracy requirements. Select appropriate control strategies, such as PID control or fuzzy control, to ensure that the changing deviations can be adapted in a dynamic environment. Based on the simulation results, optimize the correction parameters. Set the initial values ​​of the correction parameters and adjust them according to the feedback from the correction simulation. Set the initial proportional gain to 2.0 and the integral gain to 0.5. Perform multiple rounds of simulation tests, gradually adjust the correction parameters, and record the deviation values ​​and correction effects of each round until the expected deviation accuracy is achieved. Generally, the goal is to achieve a deviation accuracy of within ±0.1 mm. Implement the correction operation in the actual equipment and apply the optimized parameters for dynamic adjustment. This operation should be performed in a real-time monitoring system to obtain immediate feedback. Record each step in the correction process, including the change in the deviation value, the timestamp of the correction operation, and the parameter settings, for subsequent analysis. Record the final deviation accuracy optimization parameters, including the various control gain values ​​and their corresponding effect evaluations. Arrange the data in a standard format for subsequent application and reference. Example record: proportional gain = 1.8, integral gain = 0.4, feedback delay = 0.2 seconds, and the final deviation accuracy is ±0.05 mm. Analyze the effect of the optimized parameters and evaluate their applicability in the actual production process. Calculate the mean, variance and other statistical data of the final deviation to ensure that the optimized parameters can operate stably. Generate a report containing the simulation results of the cloth deviation correction and the secondary deviation accuracy correction process. The report should describe the parameter setting, implementation process and effect evaluation of each step in detail to provide guidance for subsequent production.

[0032] Step S6: Perform instant deviation correction delay analysis based on the cloth position deviation accuracy optimization parameters, and perform position correction iterative learning to build an intelligent cloth deviation correction control model.

[0033] In this embodiment, during the instant deviation correction process, it is first necessary to identify the delay factors that affect the deviation correction efficiency. These factors include the sensor response time, the controller calculation time, and the actuator reaction time. By installing high-precision sensors (such as laser rangefinders) and high-speed data acquisition systems in the system, the time delays of each link are recorded in real time. The sampling frequency is set to 100 Hz to ensure that the acquired delay data is detailed enough. The collected delay data is statistically analyzed, and the delay characteristics are evaluated using time domain analysis and frequency domain analysis. Important indicators such as average delay, maximum delay, and standard deviation are calculated. If the sensor response time is 50 ms on average, the controller calculation time is 30 ms, and the actuator reaction time is 20 ms, a delay model is established to take the influence of various factors into consideration. A linear regression model or a neural network model can be used for fitting to evaluate the contribution of different factors to the overall deviation correction delay. A real-time feedback mechanism is designed. According to the results of the delay analysis, the control strategy and parameter settings are adjusted to optimize the overall deviation correction performance. The feedback time interval is set to 500. ms, in order to quickly respond to dynamic changes, a dynamic compensation mechanism is introduced into the control algorithm to ensure that the compensation value can be adjusted in real time during the correction process to cope with different delay situations. Based on the existing correction accuracy optimization parameters, an iterative learning mechanism for position correction is constructed. This mechanism should be able to adjust the control strategy according to real-time feedback to achieve adaptive correction. A suitable machine learning algorithm, such as reinforcement learning or incremental learning, is selected so that parameters can be updated according to the results after each correction. The learning rate is set to 0.1. In order to balance learning and stability, in the actual production environment, collect data such as input parameters, deviation values, and compensation values ​​during each correction process for iterative learning. Set the data collection cycle to once every 1 second to ensure the real-time and effectiveness of the data. Use the collected data to train the model and update the control parameters. After each correction, use the feedback data to update the model, calculate the new compensation value, and record the results of each iteration. During the iterative learning process, regularly evaluate the performance of the model and set performance indicators such as deviation accuracy, response time, and stability to facilitate real-time monitoring of the model's effect. Use the cross-validation method to ensure the generalization ability of the model. After each iteration, calculate the performance of the model on the validation set to determine whether further adjustment of the learning parameters is needed. Or algorithm structure, integrate the delay analysis results and iterative learning mechanism into a complete intelligent control model. The model should have the ability of real-time monitoring, dynamic adjustment and adaptive deviation correction. Select a suitable control framework (such as ROS or LabVIEW) to implement the functional modules of the intelligent control model, including data acquisition module, delay analysis module, iterative learning module and control execution module. In actual operation, monitor the dynamic changes of cloth in real time, continuously adjust the deviation correction strategy through the intelligent control model, set the operation cycle to once per second, ensure that it can respond quickly to changes in cloth position, and record the results of each operation, including changes in deviation value and deviation correction response time, for subsequent analysis and optimization. Example operation results: the average deviation is reduced to ±0.05 mm, and the response time is stable within 200 ms. Analyze the operation results of the intelligent control model, evaluate its effect in actual production, calculate the percentage of deviation reduction and the degree of improvement in response time, generate a detailed report, record the parameter settings, implementation process and effect evaluation of each step, and ensure that relevant personnel can understand and apply the intelligent cloth deviation correction control model. .

[0034] In this embodiment, refer to Figure 2 , is a flowchart of detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Step S11: obtaining a real-time monitoring image of cloth on a textile device based on a high-resolution industrial camera; Step S12: performing global image brightness enhancement on the real-time cloth monitoring image to obtain an image brightness enhanced cloth image; Step S13: performing edge texture detail recognition on the cloth image with enhanced image brightness to extract the edge texture details of the cloth; Step S14: performing edge texture sharpening based on the edge texture details of the cloth, thereby generating an edge detail optimized cloth image; Step S15: performing end-to-end super-resolution reconstruction on the edge detail optimized cloth image to construct a super-resolution reconstructed cloth image.

[0035] In this embodiment, a high-resolution industrial camera (such as 1024x768 or higher resolution) is selected to ensure that the details of the cloth are captured. The camera should have a high frame rate (such as 30fps) to enable real-time monitoring on a fast-moving textile device. Configure the exposure time and gain of the camera to ensure that clear images can be obtained under different lighting conditions. In a dark environment, the exposure time needs to be set to 1 / 50 second and the gain needs to be set to 20dB. Install the industrial camera in a suitable position on the textile device to ensure that the entire surface of the cloth can be captured. Use a stable bracket to fix the camera to avoid vibration and displacement. Connect to the camera through a computer or embedded system, set image acquisition parameters (such as resolution, frame rate, and image format) to ensure that real-time images can be effectively captured and stored. Start the image acquisition system and start real-time monitoring of the cloth. Record the timestamp and related parameters of each frame of the image for subsequent analysis and processing. It is recommended to set the image saving format to TIFF to retain more image details. Use an image processing library (such as OpenCV or PIL) to load the real-time monitoring image. First, convert the image to a format suitable for processing (such as RGB) to ensure the accuracy of subsequent operations. Use histogram equalization or adaptive histogram equalization (CLAHE) to enhance the brightness of the image. Histogram equalization can effectively expand the dynamic range of the image and improve the contrast and visibility of the image. Set parameters such as the block size of CLAHE (usually set to 8x8) and the contrast limit (set to 2.0) to avoid oversaturation during the enhancement process. After processing, save the enhanced image as a new file (such as the enhanced image is named "enhanced_image.png") for subsequent use, and record the image histogram before and after enhancement for comparative analysis. Use the Canny edge detection algorithm or the Sobel operator for edge detection. The Canny algorithm has good noise resistance and accuracy, and is suitable for extracting cloth images with rich details. Set the parameters of the Canny algorithm, such as the high threshold and low threshold (the high threshold is set to 150 and the low threshold is set to 50) to effectively filter out the edges of the cloth. Use the selected edge detection algorithm to process the enhanced cloth image and generate an edge map. During this process, record the coordinate information of each edge point for subsequent analysis. For the Sobel operator, calculate the gradient magnitude and direction of the image to identify the main edge parts in the image. Visualize the extracted edge details, generate an edge image and save it for comparative analysis. Display the edge image side by side with the enhanced original image to ensure that the extraction effect can be observed intuitively. Select a suitable sharpening method, such as Laplacian operator or Unsharp Masking. UnsharpMasking can enhance the details of the image and improve the visual effect.Set the parameters of Unsharp Masking, such as blur radius (usually set to 1-2 pixels) and enhancement factor (set to 1.5 to 2.0) to ensure that the sharpening effect is natural and does not produce over-sharpening. Apply the selected sharpening algorithm to the image after edge detail recognition to enhance the contrast of the edge part of the image and make the texture of the cloth clearer. After processing, record the changes of each pixel to ensure that the details are preserved during the sharpening process. Save the sharpened image and compare it with the unsharpened image for analysis. Record the image quality indicators (such as PSNR and SSIM) before and after processing to evaluate the effectiveness of the sharpening effect. Select a suitable super-resolution reconstruction model, such as SRCNN (Super-Resolution Convolutional Neural Network) or ESPCN (Efficient Sub-Pixel Convolutional Neural Network). These two models have a high effect in processing image details. Set the input size and output size of the model, for example, set the size of the input image to 256x256 and the output to 512x512 to achieve super-resolution reconstruction. If the model you are using requires training, prepare a suitable training dataset (e.g., high-resolution cloth images and low-resolution versions) and train the model. Set training parameters such as learning rate (usually set to 0.001), batch size (e.g., 32), and number of iterations. Apply the trained super-resolution model to reconstruct the cloth image with optimized edge details to generate a high-resolution image. Evaluate the reconstructed high-resolution cloth image and record metrics such as PSNR and SSIM to quantify the quality of the reconstruction. Compare with the original high-resolution image to ensure that the reconstruction effect meets expectations. Save the final super-resolution reconstructed image and record each parameter setting during the process for subsequent analysis and quality control.

[0036] In this embodiment, the specific steps of step S15 are: Perform multi-scale convolution decomposition on the edge detail optimized cloth image to extract image convolution features of different scales; Perform principal component analysis on image convolution features of different scales to extract key convolution feature vectors of the image; A convolutional neural network architecture is used, wherein the convolutional neural network architecture includes multiple convolutional layers, pooling layers, and activation layers; The key convolution feature vector of the image is input into the convolutional neural network architecture for end-to-end deep learning to generate end-to-end convolutional visual features; Back-propagation optimization is performed on the end-to-end convolutional visual features to generate parameters that minimize the image convolution loss; Perform super-resolution iterative training on the parameters that minimize the image convolution loss to obtain image super-resolution iterative optimization parameters; The edge detail optimized cloth image is subjected to full-image super-resolution reconstruction according to the image super-resolution iterative optimization parameters to construct a super-resolution reconstructed cloth image.

[0037] In this embodiment, convolution kernels of different sizes (such as 3x3, 5x5 and 7x7) are selected to extract image features at multiple scales. Smaller convolution kernels are suitable for capturing details, while larger convolution kernels can extract more extensive contextual information. The stride and padding of the convolution operation are configured. Usually, the stride is set to 1 and the padding is set to "same" to ensure that the output image is the same size as the input image. A deep learning framework (such as TensorFlow or PyTorch) is used to implement multi-scale convolution operations. A function is defined to apply convolution kernels of different sizes to the edge detail optimized cloth image to extract the corresponding convolution feature map. The image is convolved with a 3x3 convolution kernel to generate a feature map A; a 5x5 convolution kernel is used to generate a feature map B; a 7x7 convolution kernel is used to generate a feature map C. Finally, all feature maps are spliced ​​or superimposed to form a multi-scale feature matrix. Before performing PCA, the extracted multi-scale convolution features are standardized to eliminate the dimensionality influence of different features. The Z-score standardization method subtracts the mean from each feature and divides it by the standard deviation. The PCA algorithm (implemented by the scikit-learn library) is used to set the number of principal components (select the first 10 principal components). PCA will extract the most informative feature directions by calculating the covariance matrix. Through the PCA transformation, the data in the original feature space is mapped to the new feature space to obtain the key convolution feature vector. The variance contribution rate of each principal component is recorded to evaluate its ability to explain the data set and construct a convolution. Neural network, usually includes multiple convolutional layers, ReLU activation layers and pooling layers. Convolutional layer 1: uses 32 3x3 convolution kernels, stride 1, followed by ReLU activation. Pooling layer 1: uses 2x2 maximum pooling, stride 2. Convolutional layer 2: uses 64 3x3 convolution kernels, stride 1, followed by ReLU activation. Pooling layer 2: uses 2x2 maximum pooling, stride 2. Fully connected layer: flattens the extracted features and connects them to the output layer. Compile the model using the cross entropy loss function and the Adam optimizer. Set the learning rate (such as 0.001), to ensure that the network can effectively converge, use the extracted image key convolution feature vector as input data, set the corresponding target output (such as high-resolution image), divide the training set and test set (such as 80% training, 20% test), ensure the generalization ability of the model, during the training process, use the training set to train CNN to minimize the loss function, monitor the loss value and accuracy during the training process, and use visualization tools (such as Matplotlib) to draw the training loss curve and accuracy curve, set an appropriate number of epochs (such as 50 epochs) and batch size (such as 32) to ensure that the model is fully trained, in each training iteration, calculate the output through the network forward propagation and compare it with the true target, calculate the loss value (such as mean square error or cross entropy loss), calculate the gradient through the back propagation algorithm, update the weights and biases in the network, use the optimizer (such as Adam or SGD) to perform parameter updates to minimize the loss function, record the loss value of each epoch, and ensure the training process The loss in gradually decreases, indicating that the model is learning. Use the minimized loss parameter as the initial weight to iteratively train the super-resolution image. Use the super-resolution reconstruction algorithm (such as SRCNN or ESPCN) to update the parameters in each iteration to optimize the quality of the output image. Set the number of iterations (such as 1000 times). Calculate the PSNR and SSIM of the output image in each iteration to evaluate the reconstruction quality. Record the super-resolution optimization parameters and their corresponding image quality indicators after each iteration for subsequent analysis. Use the trained super-resolution model to process the edge detail optimized cloth image in its entirety to generate a high-resolution image. Apply the optimized parameters to the reconstruction algorithm to ensure that the details and texture of the cloth are retained during the reconstruction process. Evaluate the quality of the reconstructed image, record indicators such as PSNR and SSIM, and compare them with the original high-resolution image to ensure that the reconstruction effect meets expectations. Finally, save the super-resolution reconstructed cloth image and record each parameter setting during the processing to provide a basis for subsequent analysis and quality control. .

[0038] In this embodiment, refer to Figure 3 , is a flowchart of detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Step S21: performing key cloth edge visual recognition on the super-resolution reconstructed cloth image, and marking a plurality of key cloth edge points; Step S22: performing image time-series frame decomposition on the super-resolution reconstructed cloth image to extract the cloth image frame sequence; Step S23: performing cloth dynamic optical flow analysis frame by frame on the cloth image frame sequence to generate cloth dynamic optical flow change data; Step S24: Tracking the dynamic changes of the cloth edge points based on the dynamic optical flow change data of the cloth based on multiple key cloth edge points, so as to obtain the dynamic movement trajectory of the cloth edge points.

[0039] In this embodiment, a super-resolution reconstruction technology (such as SRCNN or GANs) is used to process a low-resolution cloth image to generate a high-resolution image, ensuring that the reconstructed image has clear details and rich texture information, and determining reconstruction parameters, such as the magnification (such as 4 times) and the number of training rounds, to ensure that the generated image quality meets expectations, and preprocessing the reconstructed cloth image, including denoising, contrast enhancement and edge sharpening, etc., using Gaussian filtering and Laplacian operators in the OpenCV library to improve image quality and highlight edge features, setting image preprocessing parameters, such as the convolution kernel size of the filter (such as 3x3 or 5x5), so as to maintain edge information while denoising, and using an edge detection algorithm (such as Canny edge detection or Sobel operator) to analyze the processed image to extract edge information of the cloth. The Canny algorithm can effectively detect the edge of the cloth through double threshold processing and non-maximum suppression. Effectively identify edges, set algorithm parameters, such as low threshold and high threshold (usually set to 100 and 200) to ensure that key edges are detected, extract key cloth edge points from the detected edges, use feature extraction algorithms (such as Harris corner detection or FAST algorithm) to determine the location of key points, mark the coordinates of key edge points, and visualize them on the image to ensure the accuracy and traceability of each key point, use the cv2.VideoCapture class of the OpenCV library to read the video stream, ensure that it can be processed frame by frame, set the frame rate of the video (such as 30 frames / second) to maintain smoothness, if the cloth image is a static image, ensure that each image represents a different time point for subsequent timing analysis, use a loop structure to read the image frame by frame and save it to the specified directory, each image frame will be named in a format containing a timestamp or serial number (such as frame_001.png, frame_002.png), for subsequent processing, record the time information of each frame for timing analysis and dynamic change research, select a suitable optical flow calculation algorithm, such as Lucas-Kanade method or Farneback method, Lucas-Kanade method is suitable for small displacement, while Farneback method is suitable for large displacement scenes, set the parameters of optical flow calculation, such as window size and pyramid layers, to adapt to the characteristics of cloth images, and perform frame-by-frame optical flow calculation on the extracted image frame sequence, using cv2.calcOpticalFlowFarneback or cv2.calcOpticalFlowPyrLK function, calculates the optical flow vector between each frame, records the size and direction of each optical flow vector for subsequent analysis, integrates the optical flow vector of each frame into a data set, and forms the dynamic optical flow change data of the cloth. The data set should contain information such as the timestamp, size, and direction of each optical flow vector. Analyze the optical flow data through statistical analysis methods (such as mean and variance) to identify the dynamic change characteristics of the cloth. Use visualization tools (such as Matplotlib or OpenCV) to display the optical flow change data to facilitate understanding of the dynamic behavior of the cloth in different time periods. Draw an optical flow diagram to display the distribution and change trend of the optical flow vector. Select a set of points from the extracted key edge points as tracking targets to ensure that these points have good visibility and stability in different frames. Initialize the parameters required for the tracking algorithm, such as the tracking window size and the maximum number of iterations, to provide High tracking accuracy, use optical flow tracking algorithm (such as Lucas-Kanade optical flow method) to dynamically track the selected edge points. The algorithm will calculate the position of each edge point in the subsequent frame based on the optical flow change data, record the position change of each edge point in each frame, form the dynamic movement trajectory of the edge point, and integrate the movement trajectory of each edge point into a data set, including the coordinates and timestamp information of the edge point in each frame. The data should be stored in a format that is easy to analyze (such as CSV file), analyze the trajectory data, calculate the dynamic characteristics of the edge point such as movement speed and acceleration, use visualization tools (such as Matplotlib) to draw the dynamic movement trajectory of the edge point, and show the dynamic changes of the cloth in different time periods. The visualization results can help evaluate the motion characteristics and dynamic behavior of the cloth, generate a final report, and record the dynamic changes of each edge point and its trajectory characteristics in detail, providing a basis for subsequent research. .

[0040] In this embodiment, refer to Figure 4 , is a flowchart of detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include: Step S31: segmenting the dynamic movement trajectory of the edge point of the cloth into multiple time point trajectories, thereby obtaining the movement trajectories of the edge point at multiple time points; Step S32: Calculate the precise spatial position of the edge points based on the movement trajectories of the edge points at multiple time points to obtain the spatial positioning coordinates of the edge points at each time point; Step S33: Calculating the cloth position deviation of the edge point spatial positioning coordinates at each time point based on the preset cloth weaving standard position, and generating the cloth position deviation value at each time point; Step S34: Perform position deviation change analysis on the cloth position deviation value at each time point, thereby generating cloth position deviation trend data.

[0041] In this embodiment, the acquired edge point dynamic movement trajectory data includes the position (x, y coordinates) and timestamp information of each edge point in different time frames, to ensure the integrity and accuracy of the data, organize the data into a time series format for subsequent processing and analysis, and use a structured data format (such as CSV) for storage to ensure that each row corresponds to the edge point position at a time point, set the time window size, and select a suitable window size (for example, every 1 second or every 5 frames) according to the characteristics of the cloth movement and the sampling frequency. The window selection should be able to capture the key changes in the movement of the edge points, record the start and end time of each time window, so as to facilitate the subsequent trajectory segmentation, use a loop structure to traverse the entire trajectory data set, and segment the edge point movement trajectory according to the set time window. The data in each window will form an independent trajectory segment. For each segmented trajectory segment, the positions of its starting point and end point are calculated. And store it in a new data structure (such as a list or array), record the edge point trajectory segments segmented out of each time window, and visualize these trajectory segments to ensure the rationality and accuracy of the segmentation results, use visualization tools (such as Matplotlib) to draw trajectory diagrams, and show the movement of each time period. Set a reference coordinate system to facilitate spatial position calculations, select a fixed point of the cloth (such as the upper left corner) as the origin, establish a two-dimensional or three-dimensional coordinate system, and determine the unit of the coordinate system (such as millimeters or centimeters) to accurately calculate the position of the edge points. According to the edge point trajectory data at each time point, use simple geometric calculations to determine the spatial position of each edge point. By converting the relative coordinates of the edge points into absolute coordinates (combined with the size and position of the cloth), the position coordinates = The origin coordinates + the relative coordinates of the edge points ensure the accurate position of each edge point in space. According to the design specifications and production standards of the cloth, the standard position coordinates of each edge point are set. These standard coordinates should reflect the ideal shape and size of the cloth. The standard coordinates of each edge point are determined and stored in the data structure for subsequent calculations. For the spatial positioning coordinates of the edge points at each time point, the position deviation is calculated using the formula: deviation value = measured coordinates − standard coordinates. The calculated deviation value should be the deviation of each edge point at each time point, ensuring that the deviation value can reflect the dynamic changes of the cloth. From step S33 Organize the deviation data set obtained in the process to ensure that the deviation value and its timestamp information at each time point are complete, arrange the data in chronological order for easy analysis, set the time window for deviation analysis, such as performing statistical analysis every 5 minutes or every 10 minutes, so as to capture the dynamic changes of the deviation, use statistical analysis methods (such as mean, standard deviation and trend analysis) to analyze the deviation data, identify the change trend of the deviation, calculate the average deviation value and fluctuation range in each time window, apply regression analysis methods (such as linear regression or polynomial regression) to fit the deviation data, and identify the long-term trend of the deviation.The analysis results are organized into trend data, and the average deviation value and change rate of each time window are recorded to form a cloth position deviation trend report. Visual tools (such as Matplotlib or Seaborn) are used to draw a deviation change trend chart to intuitively show the change of cloth position deviation over time.

[0042] In this embodiment, step S4 includes the following steps: Step S41: obtaining material parameters of the current textile cloth; Step S42: performing cloth thickness identification on the material parameters of the current textile cloth to generate cloth thickness parameters; Step S43: performing fabric material analysis on the material parameters of the current textile cloth to generate textile cloth material characteristics; Step S44: calculating the cloth feeding speed of the textile device according to the cloth image frame sequence, and extracting the cloth feeding speed; Step S45: conducting real-time textile working condition mining on the cloth thickness parameter, the textile cloth material characteristics and the cloth feeding speed, thereby generating the current textile cloth working condition characteristics; Step S46: performing adaptive position deviation compensation calculation according to the cloth position deviation trend data and the current textile cloth working condition characteristics, thereby generating a cloth position deviation compensation value.

[0043] In this embodiment, suitable sensors and devices are selected to obtain material parameters of the cloth, including infrared sensors, ultrasonic sensors and stretch testers. Infrared sensors can be used to measure the thermal properties of the cloth, ultrasonic sensors can be used to measure thickness, and stretch testers can be used to obtain the strength and elastic parameters of the cloth. The calibration of the sensors is ensured to improve the measurement accuracy. The calibration process includes testing on known standard materials and recording data, collecting material parameters in real time through a data acquisition system (such as Arduino or Raspberry Pi), setting a suitable sampling frequency (such as sampling 10 times per second) to obtain sufficient real-time data, and when recording data, ensuring that the physical and chemical properties of the material are included, such as density, elastic modulus, tensile strength and thermal conductivity, etc., an ultrasonic thickness gauge or a laser rangefinder is used to accurately measure the thickness of the cloth. The ultrasonic thickness gauge calculates the thickness by emitting ultrasonic waves and measuring the reflection time, while the laser rangefinder uses the reflection of the laser beam to achieve this. The measurement position is determined. Usually, several points of the cloth are selected for multiple measurements to improve the reliability of the results. The measured thickness data is processed to calculate the average thickness value of the cloth. A simple statistical method, such as the arithmetic mean, is used to obtain the most accurate value. The final cloth thickness parameters are recorded, and the thickness value and its corresponding timestamp of each measurement point are recorded for subsequent analysis and comparison. The test results are analyzed using data analysis software (such as MATLAB or SciPy library in Python) to extract key material features, such as elastic modulus, tensile strength and tear strength, etc. The data of each feature is recorded to form a cloth material feature data set for subsequent analysis. The moving speed of the cloth between consecutive frames is calculated using the optical flow method (such as Lucas-Kanade method or Farneback method). According to the change of the optical flow vector, the moving distance of the cloth in unit time is calculated. The time interval (such as 1 second) is set as the basis for calculating the speed. Speed ​​= Displacement / time, store the calculated cloth feeding speed in the data structure for subsequent analysis, record the speed value and its corresponding timestamp at each time point, use a real-time data processing framework (such as Apache Kafka or Apache Spark Streaming) to analyze the integrated data in real time, set monitoring indicators (such as cloth thickness, material characteristics and the range of change of cloth feeding speed) to analyze the current working conditions, and generate the working condition characteristics of the current textile cloth according to the real-time analysis results, such as the overall quality evaluation of the cloth, cloth feeding stability and thickness uniformity, etc., select a suitable compensation algorithm (such as PID control algorithm) for adaptive position deviation compensation calculation, and the PID controller can be dynamically adjusted according to the current deviation, historical deviation and deviation change rate. According to the deviation trend data and the current working condition characteristics, the compensation value at each time point is calculated, and the calculated compensation value is recorded in the data structure to ensure that the compensation value at each time point and its corresponding timestamp information are complete.

[0044] In this embodiment, step S5 includes the following steps: Step S51: performing cloth deviation correction simulation on the textile device according to the cloth position deviation compensation value, and collecting cloth deviation correction simulation data; Step S52: performing secondary cloth position deviation identification on the cloth deviation correction simulation data according to the preset cloth weaving standard position, and extracting the cloth position deviation time point; Step S53: locating the position deviation edge point based on the cloth position deviation time point; Step S54: performing secondary deviation accuracy correction on the position deviation edge points, thereby generating cloth position deviation accuracy optimization parameters.

[0045] In this embodiment, the parameters of the correction simulation are determined, including the simulation time, the simulation step size, and the feedback mechanism. The simulation time is set to 10 seconds, and the position is updated once per second. According to the cloth position deviation compensation value generated in step S46, a control algorithm is written, and the compensation value is applied to the motion control of the textile device. The control algorithm uses a PID controller or a fuzzy controller to adjust the movement direction and speed of the cloth. Set motion instructions, such as "move 1 mm to the left" or "move up the calculated compensation value" to achieve accurate cloth correction. During the correction simulation process, a sensor (such as a position sensor or a camera) is used to monitor the position change of the cloth in real time. Set the sampling frequency (such as 10 times per second) to capture the dynamic changes of the cloth during the correction process. Record the cloth position at each time point and its corresponding compensation instruction to form a complete correction simulation data set. Determine the standard position of the cloth, which is usually set according to the design drawing or production specification. The standard position should include the coordinates of each edge point of the cloth in an ideal state (such as the upper left corner, the upper right corner, etc.). Store the data of the standard position in a database for subsequent comparison and analysis. According to the cloth deviation correction simulation data, the cloth position deviation at each time point is calculated using the formula, deviation = current position − standard position, and the deviation threshold (such as ±0.5 mm) is set to determine when a significant deviation occurs. When the deviation exceeds the set range, the time point is recorded. All time points at which significant deviations occur are recorded to form a deviation time point data set. Each time point should contain the deviation value and its corresponding timestamp. The extracted deviation time point data is stored in the database for subsequent analysis and processing. The cloth image is analyzed using an edge detection algorithm (such as Canny edge detection or Sobel operator) to identify edge points. Appropriate parameters are selected to ensure that the edges of the cloth can be accurately detected. For each deviation time point, the corresponding cloth image is extracted, and the edge detection algorithm is applied to generate the edge point coordinates. The edge point coordinates corresponding to each deviation time point are recorded to form an edge point data set. Each edge point should contain its coordinate value and information about the deviation time point. Select an appropriate correction method, such as the least squares method or the RANSAC algorithm, to perform precision correction on the deviation edge points. These algorithms can effectively handle noise and outliers and improve the accuracy of correction. Set the parameters of the correction model, such as the number of iterations and the convergence threshold, to ensure the stability and accuracy of the correction process. Apply the selected correction algorithm to each deviation edge point and calculate the corrected position coordinates. The correction process should take into account the previous deviation data and the actual position of the edge point. Record the corrected edge point coordinates and their corresponding deviation values ​​to form the cloth position deviation accuracy optimization parameters. Each optimization parameter should contain comparison data before and after correction to facilitate the evaluation of the correction effect.

[0046] In this embodiment, step S6 includes the following steps: Step S61: performing real-time cloth deviation correction control based on cloth position deviation accuracy optimization parameters and collecting cloth deviation correction response data; Step S62: performing instant deviation correction delay analysis on the cloth deviation correction response data to generate instant deviation correction feedback delay data; Step S63: optimizing the correction delay control on the instant correction feedback delay data to obtain correction delay optimization data; Step S64: Perform position correction iterative learning on the deviation correction delay optimization data to build an intelligent cloth deviation correction control model.

[0047] In this embodiment, a suitable control system (such as a PLC or an embedded controller) is selected to perform real-time cloth deviation correction control, ensuring that the system can receive and process input data from sensors and algorithms, and configuring the control system parameters, including the control cycle (such as 50 milliseconds) and the response time, to ensure that the deviation of the cloth can be quickly responded to. The cloth position deviation accuracy optimization parameters generated in the process are used to write a real-time control algorithm. The algorithm should include instructions for dynamically adjusting the cloth position, such as "move X mm to the left" or "adjust Y mm upwards". The control logic is set, such as using a PID controller or a fuzzy controller to adjust the cloth position in real time. The parameters of the PID controller (such as proportional, integral, and differential coefficients) should be debugged according to the actual situation. During the correction control process, sensors (such as position sensors or cameras) are used to monitor the response of the cloth in real time. The sampling frequency is set (such as 10 times per second) to capture the dynamic changes of the cloth during the correction process. The cloth position, correction instructions, and corresponding response time at each time point are recorded to form a complete response data set. The correction delay is defined as the time required from sending the correction instruction to the actual response. The delay can be calculated by recording the instruction sending time and the actual response time at each time point. The delay threshold is set to identify significant delays. For example, a delay of more than 100 milliseconds will be considered an abnormality. The correction delay at each time point is calculated: delay = actual response time - instruction sending time. The calculated instant correction feedback delay data is stored in the database to ensure that each correction can be tracked and analyzed. The deviation delay situation, each delay data should contain a timestamp, a delay value and a corresponding correction instruction, so as to facilitate subsequent analysis and processing, select a suitable optimization algorithm (such as a genetic algorithm, a particle swarm optimization or a gradient descent method) to adjust the control parameters to reduce the correction delay, the optimization algorithm should be able to automatically adjust the control parameters according to the delay data feedback, determine the optimization goal, such as minimizing the delay or maximizing the response speed, take the collected delay data as input, run the optimization algorithm to adjust the control parameters, test the effects of different parameter combinations through multiple experiments to find the best control strategy, set convergence conditions, such as stopping the optimization when the delay average value changes less than a certain threshold (such as 5 milliseconds), record the optimized control parameters and the corresponding correction delay data into the database to form a correction delay optimization data set, record the parameters of each optimization iteration and its impact on the delay, so as to facilitate subsequent analysis and verification, select a suitable machine learning model (such as a support vector machine, a random forest or a deep learning model) for iterative learning of position correction, the selected model should be able to process time series data and adapt to real-time feedback, determine the input features, such as correction parameters, delay data and the current state of the cloth, so as to facilitate model training, use step S63 The correction delay optimization data generated in the model is used for model training. The data is divided into a training set and a test set to verify the accuracy and robustness of the model. The training parameters, such as the learning rate and the number of iterations, are set.Conduct multiple trainings to optimize model performance, evaluate model performance after training, use accuracy, recall and other indicators to ensure that the model can effectively predict the deviation correction parameters, record the evaluation results in the database for subsequent optimization and improvement, apply the trained model to real-time cloth deviation correction control, adjust the deviation correction parameters in real time according to the input features to improve the deviation correction accuracy and response speed, record the feedback data after the model is applied, and perform iterative updates to continuously optimize the control strategy.

[0048] In this embodiment, a cloth weaving position correction system based on image visual recognition is provided, which is used to execute the cloth weaving position correction method based on image visual recognition as described above, including: A super-resolution reconstruction module is used to obtain a real-time monitoring image of cloth on a textile device based on a high-resolution industrial camera; the real-time monitoring image of the cloth is subjected to global image brightness enhancement and end-to-end super-resolution reconstruction to construct a super-resolution reconstructed cloth image; an edge point movement trajectory module is used to perform key cloth edge visual recognition on the super-resolution reconstructed cloth image and track the dynamic changes of edge points to obtain the dynamic movement trajectory of the cloth edge points; The position deviation calculation module is used to calculate the precise spatial position of the edge points of the cloth according to the dynamic movement trajectory of the edge points, and calculate the cloth position deviation, so as to generate cloth position deviation trend data; The deviation compensation module is used to obtain the material parameters of the current textile cloth; conduct real-time textile working condition mining on the material parameters of the current textile cloth, and perform adaptive position deviation compensation calculation based on the cloth position deviation trend data, thereby generating a cloth position deviation compensation value; The secondary deviation correction module is used to simulate the cloth deviation correction of the textile device according to the cloth position deviation compensation value, and then perform secondary deviation accuracy correction to generate cloth position deviation accuracy optimization parameters; The deviation correction control module is used to perform instant deviation correction delay analysis based on the cloth position deviation accuracy optimization parameters, and to perform iterative learning of position correction to build an intelligent cloth deviation correction control model. The present invention improves the resolution of the image through super-resolution reconstruction, making the details and edges of the cloth clearer and providing more accurate image information for subsequent processing. The high-resolution image provides more accurate data for subsequent edge recognition and position calculation, thereby improving the reliability and correction accuracy of the entire system. The movement of the cloth is accurately captured through dynamic tracking of edge points, providing key data for subsequent deviation calculation. The movement trajectory of the edge points can help the system locate the position of the cloth more accurately and detect any potential deviations in time. Through the precise calculation of the dynamic trajectory of the edge points, the trend of the cloth position deviation is monitored in real time, and the moment when the cloth deviates from the standard position is identified in advance. The generated deviation trend data will provide a detailed adjustment basis for the subsequent deviation compensation module, ensuring that the correction operation is more accurate and timely. By acquiring the working conditions and material parameters of the cloth in real time, the compensation strategy is adaptively adjusted, thereby It can adapt to the needs of cloth correction under different materials and different working conditions. Through real-time analysis of the dynamic changes of cloth, it can accurately calculate the compensation value, reduce the impact caused by material differences or working condition fluctuations, and improve the overall stability of the system. The secondary deviation correction can further fine-tune on the basis of the first correction to ensure that the accuracy of the cloth position meets the expected standard. The secondary precision correction can effectively avoid the accumulation of errors, ensure that the cloth position is always within a reasonable deviation range, and improve the overall production quality. Through delay analysis, it can evaluate and reduce the delay in the correction process and improve the real-time response of the system. Through iterative learning and adaptive adjustment, the intelligent control model can automatically optimize the correction strategy, improve the system's adaptability in different production environments, and reduce human intervention. After continuous iterative training, the correction control model can gradually improve the long-term accuracy of cloth position correction and ensure the long-term stability of the production process.

[0049] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0050] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for correcting cloth weaving position based on image visual recognition, characterized in that: The following steps are involved: Step S1: obtaining a real-time monitoring image of cloth on a textile device based on a high-resolution industrial camera; Performing global image brightness enhancement and end-to-end super-resolution reconstruction on the real-time cloth monitoring image to construct a super-resolution reconstructed cloth image; Step S2: performing visual recognition of key cloth edges on the super-resolution reconstructed cloth image, and tracking the dynamic changes of edge points, thereby obtaining the dynamic movement trajectory of the cloth edge points; Step S3: Calculate the precise spatial position of the edge points based on the dynamic movement trajectory of the edge points of the cloth, and calculate the cloth position deviation, so as to generate cloth position deviation trend data; Step S4: obtaining material parameters of the current textile cloth; performing real-time textile working condition mining on the material parameters of the current textile cloth, and performing adaptive position deviation compensation calculation according to cloth position deviation trend data, thereby generating a cloth position deviation compensation value; Step S5: performing cloth deviation correction simulation on the textile device according to the cloth position deviation compensation value, and then performing secondary deviation accuracy correction, thereby generating cloth position deviation accuracy optimization parameters; Step S6: Perform instant deviation correction delay analysis based on the cloth position deviation accuracy optimization parameters, and perform position correction iterative learning to build an intelligent cloth deviation correction control model.

2. The cloth weaving position correction method based on image visual recognition according to claim 1 is characterized in that: The specific steps of step S1 are: Step S11: obtaining a real-time monitoring image of cloth on a textile device based on a high-resolution industrial camera; Step S12: performing global image brightness enhancement on the real-time cloth monitoring image to obtain an image brightness enhanced cloth image; Step S13: performing edge texture detail recognition on the cloth image with enhanced image brightness to extract the edge texture details of the cloth; Step S14: performing edge texture sharpening based on the edge texture details of the cloth, thereby generating an edge detail optimized cloth image; Step S15: performing end-to-end super-resolution reconstruction on the edge detail optimized cloth image to construct a super-resolution reconstructed cloth image.

3. The method according to claim 2, characterized in that The specific steps of step S15 are: Perform multi-scale convolution decomposition on the edge detail optimized cloth image to extract image convolution features of different scales; Perform principal component analysis on image convolution features of different scales to extract key convolution feature vectors of the image; A convolutional neural network architecture is used, wherein the convolutional neural network architecture includes multiple convolutional layers, pooling layers, and activation layers; The key convolution feature vector of the image is input into the convolutional neural network architecture for end-to-end deep learning to generate end-to-end convolutional visual features; Back-propagation optimization is performed on the end-to-end convolutional visual features to generate parameters that minimize the image convolution loss; Perform super-resolution iterative training on the parameters that minimize the image convolution loss to obtain image super-resolution iterative optimization parameters; The edge detail optimized cloth image is subjected to full-image super-resolution reconstruction according to the image super-resolution iterative optimization parameters to construct a super-resolution reconstructed cloth image.

4. The method according to claim 1, characterized in that The specific steps of step S2 are: Step S21: performing key cloth edge visual recognition on the super-resolution reconstructed cloth image, and marking a plurality of key cloth edge points; Step S22: performing image time-series frame decomposition on the super-resolution reconstructed cloth image to extract the cloth image frame sequence; Step S23: performing cloth dynamic optical flow analysis frame by frame on the cloth image frame sequence to generate cloth dynamic optical flow change data; Step S24: Tracking the dynamic changes of the cloth edge points based on the dynamic optical flow change data of the cloth based on multiple key cloth edge points, so as to obtain the dynamic movement trajectory of the cloth edge points.

5. The cloth weaving position correction method based on image visual recognition according to claim 1 is characterized in that: The specific steps of step S3 are: Step S31: segmenting the dynamic movement trajectory of the edge point of the cloth into multiple time point trajectories, thereby obtaining the movement trajectories of the edge point at multiple time points; Step S32: Calculate the precise spatial position of the edge points based on the movement trajectories of the edge points at multiple time points to obtain the spatial positioning coordinates of the edge points at each time point; Step S33: Calculating the cloth position deviation of the edge point spatial positioning coordinates at each time point based on the preset cloth weaving standard position, and generating the cloth position deviation value at each time point; Step S34: Perform position deviation change analysis on the cloth position deviation value at each time point, thereby generating cloth position deviation trend data.

6. The cloth weaving position correction method based on image visual recognition according to claim 1 is characterized in that: The specific steps of step S4 are: Step S41: obtaining material parameters of the current textile cloth; Step S42: performing cloth thickness identification on the material parameters of the current textile cloth to generate cloth thickness parameters; Step S43: performing fabric material analysis on the material parameters of the current textile cloth to generate textile cloth material characteristics; Step S44: calculating the cloth feeding speed of the textile device according to the cloth image frame sequence, and extracting the cloth feeding speed; Step S45: conducting real-time textile working condition mining on the cloth thickness parameter, the textile cloth material characteristics and the cloth feeding speed, thereby generating the current textile cloth working condition characteristics; Step S46: performing adaptive position deviation compensation calculation according to the cloth position deviation trend data and the current textile cloth working condition characteristics, thereby generating a cloth position deviation compensation value.

7. The cloth weaving position correction method based on image visual recognition according to claim 1 is characterized in that: The specific steps of step S5 are: Step S51: performing cloth deviation correction simulation on the textile device according to the cloth position deviation compensation value, and collecting cloth deviation correction simulation data; Step S52: performing secondary cloth position deviation identification on the cloth deviation correction simulation data according to the preset cloth weaving standard position, and extracting the cloth position deviation time point; Step S53: locating the position deviation edge point based on the cloth position deviation time point; Step S54: performing secondary deviation accuracy correction on the position deviation edge points, thereby generating cloth position deviation accuracy optimization parameters.

8. The cloth weaving position correction method based on image visual recognition according to claim 1 is characterized in that: The specific steps of step S6 are: Step S61: performing real-time cloth deviation correction control based on cloth position deviation accuracy optimization parameters and collecting cloth deviation correction response data; Step S62: performing instant deviation correction delay analysis on the cloth deviation correction response data to generate instant deviation correction feedback delay data; Step S63: optimizing the correction delay control on the instant correction feedback delay data to obtain correction delay optimization data; Step S64: Perform position correction iterative learning on the deviation correction delay optimization data to build an intelligent cloth deviation correction control model.

9. A cloth weaving position correction system based on image visual recognition, characterized in that: The method for correcting the position of a piece of cloth weaving based on image visual recognition according to claim 1 comprises: A super-resolution reconstruction module is used to obtain a real-time monitoring image of cloth on a textile device based on a high-resolution industrial camera; perform global image brightness enhancement and end-to-end super-resolution reconstruction on the real-time monitoring image of the cloth to construct a super-resolution reconstructed cloth image; The edge point moving trajectory module is used to perform key cloth edge visual recognition on the super-resolution reconstructed cloth image and track the dynamic changes of the edge points, thereby obtaining the dynamic moving trajectory of the cloth edge points; The position deviation calculation module is used to calculate the precise spatial position of the edge points of the cloth according to the dynamic movement trajectory of the edge points, and calculate the cloth position deviation, so as to generate cloth position deviation trend data; The deviation compensation module is used to obtain the material parameters of the current textile cloth; conduct real-time textile working condition mining on the material parameters of the current textile cloth, and perform adaptive position deviation compensation calculation based on the cloth position deviation trend data, thereby generating a cloth position deviation compensation value; The secondary deviation correction module is used to simulate the cloth deviation correction of the textile device according to the cloth position deviation compensation value, and then perform secondary deviation accuracy correction to generate cloth position deviation accuracy optimization parameters; The deviation correction control module is used to perform instant deviation correction delay analysis based on the cloth position deviation accuracy optimization parameters, and to perform iterative learning of position correction to build an intelligent cloth deviation correction control model.

Images