Accurate processing device and processing method for clothing lock hole and nail button

Abstract

The invention discloses a garment keyhole sewing button precise processing device and method, and relates to the related technical field of garment processing, and the device comprises a keyhole feature sequence obtaining module which is used for recognizing a keyhole image sequence through a convolutional neural network, and obtaining a keyhole feature sequence; the lock hole feature sequence compensation module is used for carrying out lock hole recognition error analysis and generating recognition compensation parameters to carry out feature compensation on the lock hole feature sequence; the button position optimization analysis module is used for carrying out iterative optimization analysis on the button position; and the mechanical button sewing operation module is used for controlling button sewing equipment to perform mechanical button sewing on the target cloth according to the optimal button coordinate sequence, and performing position self-adaptive correction according to the track monitoring result. The technical problems that in the prior art, lock hole positioning is inaccurate, error compensation is insufficient, the button sewing efficiency is low, and button sewing is inaccurate due to cloth deformation are solved, and the technical effects that the garment lock hole machining precision, the button sewing efficiency and quality are improved, and the button position is optimized are achieved.

Description

Technical Field

[0001] The present application relates to the technical field related to clothing processing, and in particular to a precise processing device and method for buttonholes and buttons of clothing. Background Art

[0002] The traditional clothing production process relies on manual or semi-automated equipment, especially in button and buttonhole processing, there are still many technical challenges, such as processing errors, inaccurate buttonhole positioning, and unstable button positions, which affect the quality of the finished product and production efficiency, especially in mass production, it is difficult to ensure that the buttonholes and button positions of each garment are accurately aligned. The current buttonhole processing mostly uses mechanized means, but its inherent defects are that it is difficult to accurately identify buttonholes to meet the precision requirements, and it is impossible to make adaptive adjustments in real time according to changes in fabrics, and it is impossible to achieve more efficient and accurate clothing processing.

[0003] Therefore, in the current related technologies, there are technical problems such as inaccurate buttonhole positioning, insufficient error compensation, low button sewing efficiency, and inaccurate button sewing caused by fabric deformation. Summary of the invention

[0004] The present application solves the technical problems of inaccurate buttonhole positioning, insufficient error compensation, low buttoning efficiency and inaccurate buttoning caused by fabric deformation in the prior art by providing a precise processing device and method for buttonhole and button fastening of clothing, thereby achieving the technical effect of improving the processing accuracy of buttonholes, buttoning efficiency and quality and optimizing the button position.

[0005] The present application provides a precise processing device for buttonholes and buttons on clothing, the device comprising: a buttonhole feature sequence acquisition module, used to collect buttonhole images in target fabric, identify the buttonhole image sequence through a convolutional neural network, and obtain a buttonhole feature sequence; a buttonhole feature sequence compensation module, used to perform buttonhole recognition error analysis according to target fabric attribute information, generate recognition compensation parameters to perform feature compensation on the buttonhole feature sequence, and obtain a corrected buttonhole feature sequence; a button position optimization analysis module, used to perform iterative optimization analysis of button positions based on the corrected buttonhole feature sequence, and generate an optimal button coordinate sequence; a mechanical buttoning operation module, used to control a buttoning device to perform mechanical buttoning on the target fabric according to the optimal button coordinate sequence, and synchronously monitor the fabric movement trajectory during the buttoning operation, and perform position adaptive correction on the target fabric according to the trajectory monitoring result until the operation is completed.

[0006] In a possible implementation, the method for precise processing of buttonholes and buttons on clothing further performs the following processing: retrieving a set of sample keyhole images, and collecting keyhole features of sample keyhole images from different sample keyhole images to obtain a set of sample keyhole features, wherein the keyhole features include size, shape, and relative position; using the set of sample keyhole images and the set of sample keyhole features as supervision data, training a convolutional neural network in combination with a gradient descent algorithm until the mean square error loss function converges, obtaining a keyhole feature recognition plug-in, and using the keyhole feature recognition plug-in to recognize a keyhole image sequence, and outputting the keyhole feature sequence.

[0007] In a possible implementation, the method for accurately processing buttonholes and buttons on clothing further performs the following processing: obtaining fabric attribute information, wherein the fabric attribute information includes at least fabric thickness and fabric elasticity; taking fabric thickness and fabric elasticity as independent variables and buttonhole recognition error as dependent variable, performing correlation tracing analysis according to a sample fabric thickness set, a sample fabric elasticity set and a sample recognition error ratio set, and constructing an error analyzer, wherein the recognition error ratio includes a size error ratio, a shape error ratio and a relative position error ratio; using the error analyzer, performing buttonhole recognition error analysis according to the target fabric attribute information, outputting the target recognition error ratio, and calculating the recognition compensation parameter.

[0008] In a possible implementation, the method for precise processing of buttonholes and buttons on clothing further performs the following processing: selecting a first corrected buttonhole feature at a first position in the corrected buttonhole feature sequence, obtaining a first preset buttonhole area and randomly setting a first buttonhole coordinate; reading predetermined buttonhole parameters of the buttonhole device, simulating buttonhole sewing according to the predetermined buttonhole parameters, according to button attribute characteristics, the first corrected buttonhole feature and the first buttonhole coordinate, and evaluating to obtain a first buttonhole quality coefficient; continuing to randomly select buttonhole coordinates in the first preset buttonhole area, and performing buttonhole quality simulation analysis until a predetermined number of optimizations is reached, outputting a first optimal button coordinate, and adding it to the optimal button coordinate sequence.

[0009] In a possible implementation, the method for accurately processing buttonholes and buttons on clothing further performs the following processing: mechanically buttoning the target fabric according to the optimal button coordinate sequence, and during the buttoning operation, monitoring and collecting button images of the target fabric, performing button position error analysis based on the button images, obtaining transverse dimension errors and longitudinal dimension errors, and continuously monitoring and generating transverse dimension error sequences and longitudinal dimension error sequences; predicting transverse dimension errors at future K processing points based on the transverse dimension error sequence, obtaining K transverse prediction errors, wherein K is an integer greater than 1; predicting longitudinal dimension errors at future K processing points based on the longitudinal dimension error sequence, obtaining K longitudinal prediction errors; and performing position adaptive correction on the target fabric based on the K transverse prediction errors and the K longitudinal prediction errors.

[0010] In a possible implementation, the clothing buttonhole and button precision processing method further performs the following processing: collecting a sample transverse error sequence set at Q consecutive processing points in the historical processing record, and obtaining the transverse dimension errors at subsequent K historical processing points of different sample transverse error sequences, to obtain a sample transverse prediction error sequence set; using the sample transverse error sequence set and the sample transverse prediction error sequence set as sample data, equally dividing them into N parts, selecting them with replacement N times, constructing a first training set, iteratively selecting them N times, and obtaining N training sets; using the N training sets to train long and short-term memory networks respectively, Until convergence, N lateral error prediction branches are obtained to form a lateral error prediction channel; the number of processing points in the lateral size error sequence is obtained, if the number of processing points is greater than or equal to Q, the number of branch selections is 1, if the number of processing points is less than Q, the ratio of the number of processing points to Q is set to P, the difference between 1 and P is multiplied by N and rounded to obtain the number of branch selections; lateral error prediction branches of the branch selection number are randomly selected from the N lateral error prediction branches, and lateral size errors are predicted for the next K processing points according to the lateral size error sequence, and the K lateral prediction errors are output.

[0011] In a possible implementation, the clothing buttonhole and button precision processing method further performs the following processing: obtaining a button operation standard, wherein the button operation standard includes a transverse dimension error threshold and a longitudinal dimension error threshold; judging the K transverse prediction errors according to the transverse dimension error threshold, and if there is a situation where the transverse dimension error threshold is exceeded, locating the earliest processing point in time to obtain the earliest transverse warning processing point; judging the K longitudinal prediction errors according to the longitudinal dimension error threshold, and if there is a situation where the longitudinal dimension error threshold is exceeded, locating the earliest processing point in time to obtain the earliest longitudinal warning processing point; determining a warning processing point according to the earliest transverse warning processing point and the earliest longitudinal warning processing point, and at the warning processing point, performing orientation correction on the target fabric according to the initial orientation.

[0012] The present application also provides a method for accurately processing buttonholes and buttons on clothing, the method comprising: collecting buttonhole images in target fabric, identifying a keyhole image sequence through a convolutional neural network, and obtaining a keyhole feature sequence; performing a keyhole recognition error analysis based on target fabric attribute information, generating a recognition compensation parameter to perform feature compensation on the keyhole feature sequence, and obtaining a corrected keyhole feature sequence; performing an iterative optimization analysis of button positions based on the corrected keyhole feature sequence, and generating an optimal button coordinate sequence; controlling a buttoning device to mechanically button the target fabric according to the optimal button coordinate sequence, and synchronously monitoring the fabric movement trajectory during the buttoning operation, and performing position adaptive correction on the target fabric according to the trajectory monitoring result until the operation is completed.

[0013] The precise processing device and processing method of buttonholes and buttons on clothing proposed in this application include a buttonhole feature sequence acquisition module, which is used to identify the buttonhole image sequence through a convolutional neural network and obtain the buttonhole feature sequence; a buttonhole feature sequence compensation module, which is used to analyze the buttonhole recognition error and generate recognition compensation parameters to perform feature compensation on the buttonhole feature sequence; a button position optimization analysis module, which is used to perform iterative optimization analysis of the button position; and a mechanical buttoning operation module, which is used to control the buttoning equipment to mechanically button the target fabric according to the optimal button coordinate sequence, and perform position adaptive correction according to the trajectory monitoring results. The technical problems of inaccurate buttonhole positioning, insufficient error compensation, low buttoning efficiency, and inaccurate buttoning caused by fabric deformation in the prior art are solved, and the technical effects of improving the processing accuracy, buttoning efficiency and quality of clothing buttonholes and optimizing button positions are achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solution of the embodiment of the present disclosure, the accompanying drawings of the embodiment of the present disclosure will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the device according to the embodiment of the present application. It should be understood that the previous or following operations are not necessarily performed precisely in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0015] Figure 1 A schematic diagram of the structure of a precise processing device for buttonholes and buttons on clothing provided in an embodiment of the present application;

[0016] Figure 2 A schematic flow chart of a method for accurately processing buttonholes and buttons on clothing provided in an embodiment of the present application.

[0017] Explanation of reference numerals: buttonhole feature sequence acquisition module 10 , buttonhole feature sequence compensation module 20 , button position optimization analysis module 30 , mechanical button sewing module 40 . DETAILED DESCRIPTION

[0018] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0019] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0020] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, device, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0021] The present application embodiment provides a clothing buttonhole and button precision processing device, such as Figure 1 As shown, the device comprises:

[0022] The buttonhole feature sequence acquisition module 10 is used to collect buttonhole images in the target fabric, identify the buttonhole image sequence through a convolutional neural network, and obtain the buttonhole feature sequence.

[0023] Preferably, an image acquisition device (such as a camera, a webcam, etc.) is used to photograph the target fabric to accurately capture the image of the keyhole part on the surface of the fabric, such as photographing the details of the keyhole area of ​​the fabric during the production process through a camera, and then recognizing the keyhole image sequence through a convolutional neural network. Specifically, the keyhole image is processed by a convolutional neural network to automatically identify the keyhole area, shape, size, position and other feature information in the image. The image sequence refers to multiple frames of images that may be taken continuously, and each frame of the image contains a keyhole pattern or image information of multiple keyholes. After the convolutional neural network completes the analysis and recognition of the keyhole image, it finally outputs a keyhole feature sequence, which contains detailed information of the keyhole, such as the shape, size, edge position, hole position, etc. of the keyhole.

[0024] Furthermore, the specific configuration of the keyhole feature sequence acquisition module 10 also includes: retrieving a sample keyhole image set, and collecting keyhole features of sample keyhole images from different sample keyhole images to obtain a sample keyhole feature set, wherein the keyhole features include size, shape and relative position; using the sample keyhole image set and the sample keyhole feature set as supervision data, combining the gradient descent algorithm to train the convolutional neural network until the mean square error loss function converges, to obtain a keyhole feature recognition plug-in, and using the keyhole feature recognition plug-in to identify the keyhole image sequence, and output the keyhole feature sequence.

[0025] Preferably, a plurality of images containing keyhole patterns are retrieved from a plurality of different sources (including different fabric types, different fabric surface conditions, different buttonhole processing methods, etc.) to form a set of sample keyhole images for training a convolutional neural network so that it can learn how to identify keyholes in different samples. The sample keyhole images contain keyholes of different sizes, shapes and arrangements, which are used to provide rich training data to help the neural network understand the various keyhole features. Then, relevant keyhole features are extracted from each keyhole image, mainly including size (the size of the keyhole, including the width and height of the keyhole hole, etc.), shape (the outline of the keyhole, such as round, oval, square, etc.) and relative position (the relative position of the keyhole on the fabric, which may refer to the distance between the keyholes, the position offset or its spatial relationship with other parts of the fabric), thereby forming a set of sample keyhole features.

[0026] Preferably, a set of sample keyhole images and a set of sample keyhole features are used as input data for supervised learning of a convolutional neural network, that is, the convolutional neural network is trained using these data, wherein each keyhole image will have a corresponding label (that is, the keyhole feature corresponding to the image) as the target output of the convolutional neural network learning. Specifically, the convolutional neural network is trained in combination with a gradient descent algorithm. During the training process, the convolutional neural network (CNN) continuously adjusts the weights and biases in the network to minimize the error between the predicted output and the true keyhole feature. The gradient descent algorithm is used to optimize the neural network, and the loss function (such as the mean square error) value is gradually reduced by calculating the gradient of the error and updating the network parameters. The mean square error is a commonly used loss function to measure the difference between the network prediction value and the actual value. After continuous training and adjustment until the loss function converges, it means that the convolutional neural network has learned how to extract the correct keyhole features from the input keyhole image, and then obtain the keyhole recognition plug-in (the trained convolutional neural network model), which can output the corresponding keyhole features (such as size, shape, position, etc.) according to the input keyhole image; finally, the keyhole feature recognition plug-in is used to identify the keyhole image sequence, that is, the keyhole image sequence is input into the keyhole feature recognition plug-in for analysis and processing, and the corresponding keyhole feature sequence is output, that is, a series of specific parameters about the keyhole (such as the size, shape, relative position, etc. of each keyhole).

[0027] The buttonhole feature sequence compensation module 20 is used to perform buttonhole recognition error analysis according to target fabric property information, generate recognition compensation parameters to perform feature compensation on the buttonhole feature sequence, and obtain a corrected buttonhole feature sequence.

[0028] Preferably, the keyhole recognition error analysis is performed based on the target fabric attribute information (such as the material, thickness, elasticity, texture characteristics, color reflection characteristics, etc. of the fabric), that is, the deviation between the actual recognition result and the expected result is evaluated, and the error caused by the fabric characteristics is found, such as the blurred keyhole boundary caused by the complex fabric texture, the deformation of the keyhole shape due to the elasticity or stretching of the fabric, and then the recognition compensation parameters are generated, wherein the recognition compensation parameters refer to the adjustment parameters generated to reduce the error obtained through historical data training and machine learning model prediction, which may include size compensation parameters (used to adjust the keyhole size change caused by stretching or scaling), shape compensation parameters (used to correct the keyhole boundary recognition deviation) and position compensation parameters. Parameters (used to correct the offset of the identified buttonhole position); then the generated error recognition compensation parameters are used to perform feature compensation on the buttonhole feature sequence (the set of buttonhole information output by the convolutional neural network recognition), that is, to adjust the feature data in the buttonhole feature sequence. For example, if the recognition error shows that the width of the buttonhole is generally smaller than the true value, the width feature is uniformly amplified for compensation. If the center position of the buttonhole is offset from the expected position, the compensation parameters are used to correct the offset of its coordinate value to ensure that the buttonhole feature is closer to the actual situation of actual fabric processing, thereby generating a corrected buttonhole feature sequence with higher accuracy and consistency, thereby significantly improving the accuracy and effect of subsequent operations (such as mechanical buttoning).

[0029] Furthermore, the specific configuration of the keyhole feature sequence compensation module 20 also includes obtaining fabric attribute information, wherein the fabric attribute information includes at least fabric thickness and fabric elasticity; taking fabric thickness and fabric elasticity as independent variables, taking keyhole recognition error as dependent variable, performing correlation tracing analysis according to a sample fabric thickness set, a sample fabric elasticity set and a sample recognition error ratio set, and constructing an error analyzer, wherein the recognition error ratio includes a size error ratio, a shape error ratio and a relative position error ratio; using the error analyzer, performing keyhole recognition error analysis according to the target fabric attribute information, outputting the target recognition error ratio, and calculating the recognition compensation parameter.

[0030] Preferably, fabric attribute information is obtained, wherein the fabric attribute information is a key factor affecting the keyhole recognition accuracy, and mainly includes two attributes: fabric thickness and fabric elasticity. Fabric thickness refers to the physical size of the fabric in the vertical direction, usually in millimeters (mm). The thickness of the fabric directly affects the mechanical behavior during the buttonhole processing process, such as the resistance when threading the needle and the appearance of the buttonhole shape on the fabric. Fabric elasticity refers to the deformation ability of the fabric when subjected to external force. Fabric with good elasticity will cause changes in the shape and position of the buttonhole due to stretching or deformation, thereby affecting the recognition accuracy. Using the two independent variables of fabric thickness and fabric elasticity, an error analyzer is constructed by performing association tracing analysis on historical sample data to predict the keyhole recognition error based on fabric attributes. Specifically, a sample fabric thickness set, a sample fabric elasticity set and a sample recognition error ratio set are collected from a large number of historical sample data of different fabrics to establish a relationship between the error and the fabric attribute. Each sample includes the thickness and elasticity of the fabric, as well as the keyhole recognition error identified by the convolutional neural network.

[0031] Preferably, through the correlation tracing analysis, the correlation between the thickness and elasticity of the fabric and the buttonhole recognition error is found, and then an error analyzer is constructed to predict the recognition error of the new target fabric; wherein the recognition error ratio includes the size error ratio (the ratio of the size deviation of the buttonhole to the expected size), the shape error ratio (the deviation ratio between the shape of the buttonhole, such as circle, ellipse, square, etc., and the actual shape, and the fabric with strong elasticity may cause the edge of the buttonhole to be irregular) and the relative position error ratio (the position offset ratio of the buttonhole relative to other parts of the fabric, and the elasticity, thickness or physical unevenness of the fabric may cause the position of the buttonhole to be offset); the target fabric attribute information (the thickness and elasticity of the fabric) is input as input data into the error analyzer for keyhole recognition error analysis, that is, predicting the possible errors in keyhole recognition, including the size error of the keyhole, the shape error of the keyhole and the position error of the keyhole, and then outputting the target recognition error ratio according to the result of the attribute output of the target fabric; then calculating the recognition error compensation parameters, including the size compensation parameters, the shape compensation parameters and the position compensation parameters, which are used to compensate for the keyhole error, so that the size, shape and position of the keyhole are more accurate.

[0032] Preferably, the size compensation parameters are used to correct the size error to ensure that the actual size of the keyhole is consistent with the predetermined size. For example, if the error analyzer predicts a size error of 10%, the compensation parameters will adjust the recognition result to ensure the accuracy of the keyhole size. The shape compensation parameters are used to correct the shape error so that the shape of the keyhole meets the design requirements. If the elasticity of the fabric causes the keyhole to deform, the compensation parameters will adjust the recognition result of the keyhole to restore the predetermined shape. The position compensation parameters are used to correct the position error to ensure the accuracy of the keyhole position. For example, if the error analyzer outputs a position error of 5mm, the compensation parameters will adjust the coordinate value of the keyhole to accurately match the predetermined position. By calculating the compensation parameters to correct these errors, the accuracy of the buttonhole processing is ensured, and the error caused by the physical properties of the fabric (such as thickness and elasticity) is avoided, thereby improving the quality and processing accuracy of the finished product.

[0033] The button position optimization analysis module 30 is used to perform iterative optimization analysis of button positions based on the corrected buttonhole feature sequence to generate an optimal button coordinate sequence.

[0034] Preferably, an iterative optimization analysis of the button position is performed according to the corrected keyhole feature sequence, that is, based on the known keyhole features, the best button position is found through continuous adjustment and calculation. Specifically, the coordinates of the initial position of the button are calculated according to the coordinates of the center point of the keyhole and the overall layout of the fabric. For example, the button usually needs to be located on the symmetry axis of the center of the keyhole, and then the optimization target is determined to ensure the alignment accuracy between the button and the keyhole while meeting the processing requirements. For example, the position of the button needs to be located at the best point in the center of the keyhole to ensure visual effect and mechanical strength. In the case of fabric elasticity or position offset, it is necessary to ensure that the button can be naturally reset to the ideal position after processing; then an iterative algorithm (such as gradient descent, genetic algorithm, etc.) is used to continuously adjust the candidate position of the button, and according to the position The accuracy (distance from the center of the buttonhole), smoothness of the processing trajectory, overall arrangement symmetry, etc. are used to calculate the fitness of each candidate position. Through iterative optimization of the button position, the best button coordinate sequence is output, which contains a set of accurate button coordinate data for guiding the operation of the mechanical button sewing equipment. Specifically, the position coordinates of each button on the fabric (such as X and Y coordinates on a two-dimensional plane) and the arrangement order of the coordinates, such as the sequential positions of a row of buttons. Among them, the best button coordinate sequence can ensure that the position of each button is accurately aligned with the buttonhole to avoid button installation deviation. At the same time, the characteristics of the fabric (such as elasticity and deformation) are taken into account, so that the button sewing effect remains in the best position when the fabric returns to its natural state, thereby significantly improving the accuracy and efficiency of the button sewing process and ensuring that the button installation meets the design requirements.

[0035] Furthermore, the specific configuration of the button position optimization analysis module 30 also includes selecting a first correction buttonhole feature at a first position in the correction buttonhole feature sequence, obtaining a first preset buttoning area and randomly setting a first buttoning coordinate; reading predetermined buttoning parameters of the buttoning device, simulating buttoning according to the predetermined buttoning parameters, according to button attribute characteristics, the first correction buttonhole feature and the first buttoning coordinate, and evaluating to obtain a first buttoning quality coefficient; continuing to randomly select buttoning coordinates in the first preset buttoning area, and performing buttoning quality simulation analysis until a predetermined number of optimizations is reached, outputting a first optimal button coordinate, and adding it to the optimal button coordinate sequence.

[0036] Preferably, a correction keyhole feature at a position is randomly selected from the correction keyhole feature sequence as the correction keyhole feature (detailed data of the keyhole after error compensation, including the keyhole position coordinates, the shape and size of the keyhole, and the relative position relationship of the keyhole on the fabric), and a first preset buttoning area is obtained, that is, a predetermined area around the first correction keyhole feature, indicating the effective range in which the buttoning operation is allowed. The size and shape of this area can be set according to the characteristics of the keyhole (such as size, shape) and the properties of the fabric (such as elasticity, thickness), usually a rectangular or circular area with the center of the keyhole as the reference point, and then a point is randomly selected in the preset buttoning area as the initial buttoning position, and a plurality of candidate coordinates are randomly selected to generate in the area, and the optimal position is found through simulation analysis and optimization algorithm.

[0037] Preferably, the predetermined buttoning parameters of the buttoning device are read. The buttoning parameters refer to key process indicators that affect processing accuracy and quality, including punching force (the pressure applied by the mechanical device when buttoning, which affects the fixing strength of the button and the fabric), thread tension (the tension maintained by the sewing thread during the buttoning process, which determines the firmness of the button), stitch spacing (the distance between the stitches of the stitching thread, which affects the stability and aesthetics of the buttoning), and needle bar running speed (the moving speed of the needle bar of the device during the buttoning process, which affects the efficiency and quality). The buttoning device is used in clothing production. Mechanical equipment for fixing buttons to fabrics, passing buttons through fabrics through specific processes, and firmly fixing buttons on fabrics by sewing, stamping or crimping, etc., mainly including workbench, needle bar mechanism (controlling the up and down movement of needle), sewing device (passing suture through fabrics and fixing buttons, including needle, thread reel, needle plate, needle and thread tension device, etc.), stamping mechanism (such as using a press to stamp buttons through fabrics), control system (PLC controller, servo motor and sensor, etc.), feeding device (precise fabric feeding), and position calibration device.

[0038] Preferably, based on the correction features of the buttonhole, the randomly generated button sewing coordinates, the button attribute features (such as button size, material) and the button sewing equipment parameters, the button sewing simulation is performed, that is, the working process of the button sewing equipment is simulated by computer virtual simulation, and its button sewing effect is evaluated. Specifically, the predetermined button sewing parameters (such as punching force, stitch spacing, etc.), button attribute features, first correction buttonhole features and first button sewing coordinates are input into the simulation model, and the actual operation process of the button sewing equipment under these parameters is simulated, including how the needle bar of the button sewing equipment passes through the cloth and the buttonhole, how the cloth deforms during the button sewing process, and how the suture passes through the buttonhole and button to form a fixed shape, thereby generating multiple simulation results, such as whether the fixed position of the button is aligned with the buttonhole, the firmness of the button sewing (such as the button The simulation results are evaluated and quantified to obtain the corresponding first button sewing quality coefficient, which usually includes a comprehensive evaluation of button sewing accuracy (whether the button is accurately positioned at the center of the buttonhole), firmness (whether the strength of the button and the fabric meets the requirements), aesthetics (whether the overall visual effect after button sewing meets the requirements, such as uniform stitches) and fabric protection (whether the fabric is damaged during the button sewing process, such as tearing or excessive deformation). The higher the quality coefficient, the more suitable the current combination of button sewing parameters, buttonhole features and button sewing coordinates is for actual production, so that the optimal button sewing position and process parameters can be found without actual processing, reducing the cost of trial and error and improving the efficiency and quality of button sewing.

[0039] Preferably, within the first preset buttoning area, new buttoning coordinates continue to be randomly selected, buttoning and quality evaluation are repeatedly simulated, and genetic algorithms, particle swarm optimization or gradient descent are used to gradually find the buttoning position with the highest quality coefficient through multiple iterative analysis (reaching a predetermined number of optimization times), and then the first optimal buttoning coordinates within the preset area are output, that is, the optimal button position under the current keyhole characteristics, which can ensure that the buttoning operation achieves the best effect in terms of firmness, accuracy and aesthetics, and the calculated first optimal buttoning coordinates are added to the overall optimal button coordinate sequence. The optimal button coordinate sequence is a collection of optimal button positions under all keyhole positions, which ultimately guides the buttoning equipment to complete accurate and efficient buttoning operations, which not only ensures the accuracy of the button position, but also optimizes production efficiency and finished product quality.

[0040] The mechanical button sewing operation module 40 is used to control the button sewing equipment to mechanically sew buttons on the target fabric according to the optimal button coordinate sequence, and synchronously monitor the fabric movement trajectory during the button sewing operation, and perform position adaptive correction on the target fabric according to the trajectory monitoring result until the operation is completed.

[0041] Preferably, the coordinates of the optimal button coordinate sequence are applied to the button-stitching device, and the button-stitching device accurately installs each button at a specified position on the fabric according to these coordinates. Specifically, based on the optimal button coordinate sequence, the button-stitching device performs the button-stitching operation through precise control, including automatically identifying the current state and position of the fabric to ensure that the fabric is fixed in the button-stitching area. According to the optimal button coordinates, the button-stitching device passes the button through the buttonhole of the fabric to accurately complete the sewing or crimping, thereby ensuring the quality of the finished garment and avoiding button deviation or fabric damage. At the same time, the fabric movement trajectory is synchronously monitored during the button-stitching operation. Specifically, the fabric may undergo slight movement changes (such as stretching, sliding or deformation) during the button-stitching process. The fabric movement is monitored in real time by integrating sensors, cameras or laser scanning devices, which work synchronously with the button-stitching device to capture fabric position, shape, speed and other information in real time. Update the motion state of the fabric, obtain the trajectory monitoring results, and adjust the button sewing operation according to the trajectory monitoring results, that is, perform position adaptive correction on the target fabric, including the button sewing equipment dynamically adjusting the position of the fabric according to the trajectory monitoring results. For example, if the fabric is offset, the equipment will automatically adjust the positioning to ensure that the buttonhole is aligned with the button. By dynamically adjusting the fabric position, it is ensured that each button sewing is performed accurately, and the button can be fixed in the best position even when the fabric moves or changes; until the job is completed, that is, after the button sewing equipment completes the sewing of each button, it automatically determines whether the target position has been reached. If so, it enters the next button sewing position; if not, it will continue to adjust until it reaches the correct position, until all buttons are sewn and the fabric is processed, indicating that the job is completed, making the button sewing process not only efficient but also precise, ensuring the high quality and consistency of the final garment product.

[0042] Furthermore, the specific configuration of the mechanical button sewing operation module 40 also includes mechanically sewing buttons on the target fabric according to the optimal button coordinate sequence, and during the button sewing operation process, monitoring and collecting button sewing images of the target fabric, performing button sewing position error analysis based on the button sewing images, obtaining transverse dimension errors and longitudinal dimension errors, and continuously monitoring and generating transverse dimension error sequences and longitudinal dimension error sequences; predicting transverse dimension errors at future K processing points based on the transverse dimension error sequence, and obtaining K transverse prediction errors, wherein K is an integer greater than 1; predicting longitudinal dimension errors at future K processing points based on the longitudinal dimension error sequence, and obtaining K longitudinal prediction errors; and performing position adaptive correction on the target fabric based on the K transverse prediction errors and K longitudinal prediction errors.

[0043] Preferably, the target fabric is mechanically buttoned according to the optimal button coordinate sequence, that is, the buttoning device accurately installs each button to a designated position on the fabric according to the optimal button coordinate sequence to ensure that each button is accurately positioned on the fabric, and an image acquisition system (such as a camera or a sensor) installed on the buttoning device is used to monitor the fabric image during the buttoning process in real time, that is, the buttoning image of the target fabric, to capture the alignment of the button position and the buttonhole on the fabric surface, that is, to obtain the actual position of the button after buttoning, and then perform button position error analysis based on the button image, that is, analyze the actual position of the button to determine the error. The deviation between the button and the optimal predetermined position is determined to obtain the lateral size error and the longitudinal size error, wherein the lateral size error refers to the deviation of the button in the horizontal direction, that is, the error of the button's position relative to the predetermined coordinate in the lateral (X-axis) direction; the longitudinal size error refers to the deviation of the button in the vertical direction, that is, the error of the button's position relative to the predetermined coordinate in the longitudinal (Y-axis) direction; and then two corresponding error sequences are generated through continuous real-time monitoring, namely, the lateral size error sequence (recording the change of the lateral error in each buttoning operation) and the longitudinal size error sequence (recording the change of the longitudinal error in each buttoning operation).

[0044] Preferably, the historical error data (i.e., the lateral dimension error sequence) is used to predict the lateral errors of the future K processing points through data analysis and prediction models (such as regression analysis, time series prediction or machine learning algorithm), that is, to obtain K lateral prediction errors, wherein K is an integer greater than 1, representing the total number of button sewing processing points where errors will occur in the future; similarly, based on the historical error data (i.e., the longitudinal dimension error sequence), the corresponding prediction model is used to predict the longitudinal errors of the future K processing points, that is, to obtain K longitudinal prediction errors, which helps to understand the longitudinal deviations that may occur in the future production process; finally, the position of the fabric is adjusted according to the predicted values ​​of the lateral and longitudinal errors to ensure that each button sewing operation can be as accurate as possible. For example, based on the predicted lateral error, the lateral position of the fabric is adjusted to ensure that the button can be accurately located at the center of the buttonhole during the next button sewing operation. Similarly, based on the predicted longitudinal error, the longitudinal position of the fabric is adjusted to avoid the longitudinal offset affecting the button sewing accuracy, thereby ensuring the stability and consistency of the button sewing accuracy throughout the production process, greatly improving the accuracy, efficiency and quality of finished products of automated production.

[0045] Furthermore, the specific configuration of the mechanical button fastening operation module 40 also includes collecting a sample lateral error sequence set at Q consecutive processing points in the historical processing record, and obtaining the lateral size errors at subsequent K historical processing points of different sample lateral error sequences to obtain a sample lateral prediction error sequence set; using the sample lateral error sequence set and the sample lateral prediction error sequence set as sample data, dividing them into N parts with equal value, selecting them N times with replacement, constructing a first training set, iteratively selecting them N times, and obtaining N training sets; using the N training sets to train the long short-term memory network respectively until convergence, N lateral error prediction branches are harvested to form a lateral error prediction channel; the number of processing points in the lateral size error sequence is obtained, if the number of processing points is greater than or equal to Q, the number of branch selections is 1, if the number of processing points is less than Q, the ratio of the number of processing points to Q is set to P, the difference between 1 and P is multiplied by N and rounded to obtain the number of branch selections; lateral error prediction branches of the branch selection number are randomly selected from the N lateral error prediction branches, and lateral size error predictions are performed at future K processing points according to the lateral size error sequence, and the K lateral prediction errors are output.

[0046] Preferably, the lateral dimension errors of Q consecutive processing points are extracted from the historical processing records to form a sample lateral error sequence set, which reflects the changing trend of the lateral error during the button processing process and serves as the input data for model prediction. At the same time, the actual lateral dimension errors of these sample lateral error sequences at the subsequent K processing points are obtained to form a sample lateral prediction error sequence set as the target value (prediction output) of the model prediction, wherein Q is an integer greater than 1, indicating the length of the historical error sequence (the total number of button processing points). Then, the sample data (sample lateral error sequence set and prediction error sequence set) are equally divided into N parts to increase the diversity of the data and the robustness of the training. The first training set is constructed by randomly selecting N times with replacement, and then iteratively selecting N times to obtain N training sets, each of which includes an input (historical error sequence) and an output (corresponding prediction error sequence), wherein N is a positive integer, which helps the model cope with the randomness and diversity of the error data and improves the prediction accuracy and generalization ability.

[0047] Preferably, N training sets are used to train long short-term memory networks respectively, wherein the long short-term memory network (LSTM) is a deep learning model for processing time series data, which is good at capturing long-term and short-term dependencies in sequence data, and is used to learn the changing pattern of lateral errors, thereby predicting future errors. Specifically, each training set is used to train an independent long short-term memory network model until the model converges (that is, the prediction error is small enough), and then N lateral error prediction branches are generated according to the training of the N training sets, which are responsible for learning and predicting errors based on different sample data, and the N lateral error prediction branches are combined into a lateral error prediction channel for lateral error prediction.

[0048] Preferably, the number of processing points in the lateral dimension error sequence is obtained, wherein the more data in the lateral dimension error sequence, the more reliable it is, the higher the prediction accuracy, the fewer branches are selected, the analysis efficiency is improved, and the computing power is saved; conversely, the less data, the more branches are selected, that is, if the number of processing points in the current lateral dimension error sequence is ≥Q (that is, the historical data is long enough), only one branch needs to be selected for prediction, and if the number of processing points is <Q (that is, the historical data is short), the number of selected branches needs to be adjusted according to the proportion. Specifically, the ratio of the number of processing points to Q is taken as the proportion P, and 1-P is used to represent the insufficient proportion, which is multiplied by N and rounded to obtain the number of branches that need to be additionally selected; finally, a lateral error prediction branch with the calculated number of branch selections is randomly selected from the N lateral error prediction branches as a candidate lateral error prediction branch, and the candidate lateral error prediction branch is used to predict the lateral dimension errors of the future K processing points according to the current lateral dimension error sequence, and the corresponding K lateral prediction errors are output, each error value corresponds to the lateral dimension error of a future processing point, thereby improving the accuracy of error prediction and production efficiency.

[0049] Furthermore, the specific configuration of the mechanical button sewing operation module 40 also includes obtaining a button sewing operation standard, wherein the button sewing operation standard includes a transverse dimension error threshold and a longitudinal dimension error threshold; judging the K transverse prediction errors according to the transverse dimension error threshold, and if there is a situation where the transverse dimension error threshold is exceeded, the earliest processing point in time is located to obtain the earliest transverse warning processing point; judging the K longitudinal prediction errors according to the longitudinal dimension error threshold, and if there is a situation where the longitudinal dimension error threshold is exceeded, the earliest processing point in time is located to obtain the earliest longitudinal warning processing point; determining a warning processing point according to the earliest transverse warning processing point and the earliest longitudinal warning processing point, and at the warning processing point, performing orientation correction on the target fabric according to the initial orientation.

[0050] Preferably, the button nailing operation standard is obtained, that is, the transverse dimension error threshold and the longitudinal dimension error threshold, wherein the transverse dimension error threshold refers to the maximum allowable deviation of the transverse position error in the button nailing operation, and the longitudinal dimension error threshold refers to the maximum allowable deviation of the longitudinal position error in the button nailing operation; then a judgment is made based on the transverse dimension error threshold, that is, if a certain predicted transverse error exceeds the preset transverse dimension error threshold, it means that the button nailing of the processing point may have a position deviation, and needs to be adjusted, and then the processing point where the transverse error exceeds the standard first is located, that is, the processing point where the error exceeds the threshold for the first time, as the transverse warning processing point; similarly, a judgment is made based on the longitudinal dimension error threshold, that is, if a certain longitudinal predicted error exceeds the set longitudinal dimension error threshold, it means that the button nailing position of the processing point may also deviate, and needs to be adjusted, and then the processing point where the longitudinal error occurs first is located. The processing point where the error exceeds the standard, that is, the longitudinal processing point where the error exceeds the threshold for the first time, is used as the longitudinal warning processing point; then, according to the earliest horizontal warning processing point and the earliest longitudinal warning processing point, that is, the earliest horizontal warning processing point and the earliest longitudinal warning processing point are compared, and the earlier one is determined as the final warning processing point; finally, under the warning processing point, the target fabric is oriented according to the initial orientation to ensure the accuracy of the button sewing operation, which specifically includes manual correction (the operator manually adjusts the position of the fabric to ensure that the fabric returns to the standard direction and standard position) and mechanical correction (the mechanical equipment automatically adjusts the position of the fabric to ensure that the fabric returns to the standard direction and standard position). After correction, the direction and position of the fabric should be restored to the preset standard state (standard direction and position) to ensure the accuracy and consistency of the button sewing operation, greatly reduce the button sewing deviation, and improve production efficiency and product quality.

[0051] In the above, refer to Figure 1 The present invention is described in detail about the precise processing device for buttonholes and buttons of clothing according to the embodiment of the present invention. Figure 2 The invention describes a method for accurately processing buttonholes and buttons on clothing according to an embodiment of the invention.

[0052] Precision processing method for buttonholes of clothing, such as Figure 2 As shown, the method includes: collecting keyhole images in the target fabric, identifying the keyhole image sequence through a convolutional neural network, and obtaining a keyhole feature sequence; performing keyhole recognition error analysis according to the target fabric attribute information, generating recognition compensation parameters to perform feature compensation on the keyhole feature sequence, and obtaining a corrected keyhole feature sequence; performing iterative optimization analysis of button positions based on the corrected keyhole feature sequence to generate an optimal button coordinate sequence; controlling the buttoning equipment to mechanically button the target fabric according to the optimal button coordinate sequence, and synchronously monitoring the fabric movement trajectory during the buttoning operation, and performing position adaptive correction on the target fabric according to the trajectory monitoring result until the operation is completed.

[0053] In a possible implementation, the method for precise processing of buttonholes and buttons on clothing further performs the following processing: retrieving a set of sample keyhole images, and collecting keyhole features of sample keyhole images from different sample keyhole images to obtain a set of sample keyhole features, wherein the keyhole features include size, shape, and relative position; using the set of sample keyhole images and the set of sample keyhole features as supervision data, training a convolutional neural network in combination with a gradient descent algorithm until the mean square error loss function converges, obtaining a keyhole feature recognition plug-in, and using the keyhole feature recognition plug-in to recognize a keyhole image sequence, and outputting the keyhole feature sequence.

[0054] In a possible implementation, the method for accurately processing buttonholes and buttons on clothing further performs the following processing: obtaining fabric attribute information, wherein the fabric attribute information includes at least fabric thickness and fabric elasticity; taking fabric thickness and fabric elasticity as independent variables and buttonhole recognition error as dependent variable, performing correlation tracing analysis according to a sample fabric thickness set, a sample fabric elasticity set and a sample recognition error ratio set, and constructing an error analyzer, wherein the recognition error ratio includes a size error ratio, a shape error ratio and a relative position error ratio; using the error analyzer, performing buttonhole recognition error analysis according to the target fabric attribute information, outputting the target recognition error ratio, and calculating the recognition compensation parameter.

[0055] In a possible implementation, the method for precise processing of buttonholes and buttons on clothing further performs the following processing: selecting a first corrected buttonhole feature at a first position in the corrected buttonhole feature sequence, obtaining a first preset buttonhole area and randomly setting a first buttonhole coordinate; reading predetermined buttonhole parameters of the buttonhole device, simulating buttonhole sewing according to the predetermined buttonhole parameters, according to button attribute characteristics, the first corrected buttonhole feature and the first buttonhole coordinate, and evaluating to obtain a first buttonhole quality coefficient; continuing to randomly select buttonhole coordinates in the first preset buttonhole area, and performing buttonhole quality simulation analysis until a predetermined number of optimizations is reached, outputting a first optimal button coordinate, and adding it to the optimal button coordinate sequence.

[0056] In a possible implementation, the method for accurately processing buttonholes and buttons on clothing further performs the following processing: mechanically buttoning the target fabric according to the optimal button coordinate sequence, and during the buttoning operation, monitoring and collecting button images of the target fabric, performing button position error analysis based on the button images, obtaining transverse dimension errors and longitudinal dimension errors, and continuously monitoring and generating transverse dimension error sequences and longitudinal dimension error sequences; predicting transverse dimension errors at future K processing points based on the transverse dimension error sequence, obtaining K transverse prediction errors, wherein K is an integer greater than 1; predicting longitudinal dimension errors at future K processing points based on the longitudinal dimension error sequence, obtaining K longitudinal prediction errors; and performing position adaptive correction on the target fabric based on the K transverse prediction errors and the K longitudinal prediction errors.

[0057] In a possible implementation, the clothing buttonhole and button precision processing method further performs the following processing: collecting a sample transverse error sequence set at Q consecutive processing points in the historical processing record, and obtaining the transverse dimension errors at subsequent K historical processing points of different sample transverse error sequences, to obtain a sample transverse prediction error sequence set; using the sample transverse error sequence set and the sample transverse prediction error sequence set as sample data, equally dividing them into N parts, selecting them with replacement N times, constructing a first training set, iteratively selecting them N times, and obtaining N training sets; using the N training sets to train long and short-term memory networks respectively, Until convergence, N lateral error prediction branches are obtained to form a lateral error prediction channel; the number of processing points in the lateral size error sequence is obtained, if the number of processing points is greater than or equal to Q, the number of branch selections is 1, if the number of processing points is less than Q, the ratio of the number of processing points to Q is set to P, the difference between 1 and P is multiplied by N and rounded to obtain the number of branch selections; lateral error prediction branches of the branch selection number are randomly selected from the N lateral error prediction branches, and lateral size errors are predicted for the next K processing points according to the lateral size error sequence, and the K lateral prediction errors are output.

[0058] In a possible implementation, the clothing buttonhole and button precision processing method further performs the following processing: obtaining a button operation standard, wherein the button operation standard includes a transverse dimension error threshold and a longitudinal dimension error threshold; judging the K transverse prediction errors according to the transverse dimension error threshold, and if there is a situation where the transverse dimension error threshold is exceeded, locating the earliest processing point in time to obtain the earliest transverse warning processing point; judging the K longitudinal prediction errors according to the longitudinal dimension error threshold, and if there is a situation where the longitudinal dimension error threshold is exceeded, locating the earliest processing point in time to obtain the earliest longitudinal warning processing point; determining a warning processing point according to the earliest transverse warning processing point and the earliest longitudinal warning processing point, and at the warning processing point, performing orientation correction on the target fabric according to the initial orientation.

[0059] Although the present application makes various references to certain modules in the device according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0060] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A precise processing device for buttonholes and buttons of clothing, characterized in that: The clothing buttonhole and button precision processing device comprises: A keyhole feature sequence acquisition module is used to collect keyhole images in the target fabric, identify the keyhole image sequence through a convolutional neural network, and obtain the keyhole feature sequence; A buttonhole feature sequence compensation module is used to analyze the buttonhole recognition error according to the target fabric property information, generate recognition compensation parameters to perform feature compensation on the buttonhole feature sequence, and obtain a corrected buttonhole feature sequence; A button position optimization analysis module, used to perform iterative optimization analysis of button positions based on the corrected buttonhole feature sequence to generate an optimal button coordinate sequence; The mechanical button sewing operation module is used to control the button sewing equipment to mechanically sew buttons on the target fabric according to the optimal button coordinate sequence, and synchronously monitor the fabric movement trajectory during the button sewing operation, and perform position adaptive correction on the target fabric according to the trajectory monitoring result until the operation is completed.

2. The clothing buttonhole precision processing device according to claim 1 is characterized in that: The keyhole feature sequence acquisition module comprises: A sample keyhole feature set acquisition unit is used to retrieve and acquire a sample keyhole image set, and collect keyhole features of sample keyhole images from different sample keyhole images to acquire a sample keyhole feature set, wherein the keyhole features include size, shape and relative position; The keyhole feature sequence output unit is used to train the convolutional neural network with the sample keyhole image set and the sample keyhole feature set as supervision data in combination with the gradient descent algorithm until the mean square error loss function converges, thereby obtaining a keyhole feature recognition plug-in, and using the keyhole feature recognition plug-in to recognize the keyhole image sequence and output the keyhole feature sequence.

3. The clothing buttonhole precision processing device according to claim 1, characterized in that: The keyhole feature sequence compensation module comprises: A cloth property information acquisition unit, used to acquire cloth property information, wherein the cloth property information at least includes cloth thickness and cloth elasticity; The error analyzer construction unit is used to take the fabric thickness and fabric elasticity as independent variables and the buttonhole recognition error as the dependent variable, and to perform correlation tracing analysis according to the sample fabric thickness set, the sample fabric elasticity set and the sample recognition error ratio set to construct an error analyzer, wherein the recognition error ratio includes the size error ratio, the shape error ratio and the relative position error ratio; The buttonhole recognition error analysis unit is used to use the error analyzer to perform buttonhole recognition error analysis according to target fabric attribute information, output a target recognition error ratio, and calculate the recognition compensation parameter.

4. The clothing buttonhole precision processing device according to claim 1, characterized in that: The button position optimization analysis module includes: A first button-stitching coordinate setting unit, configured to select a first correction buttonhole feature at a first position in the correction buttonhole feature sequence, obtain a first preset button-stitching area and randomly set a first button-stitching coordinate; A button sewing simulation unit is used to read predetermined button sewing parameters of a button sewing device, simulate button sewing according to the predetermined button sewing parameters, the button attribute characteristics, the first correction buttonhole characteristics and the first button sewing coordinates, and evaluate and obtain a first button sewing quality coefficient; The first optimal button coordinate output unit is used to continue to randomly select button coordinates in the first preset button area and perform button quality simulation analysis until a predetermined number of optimization times is reached, output the first optimal button coordinates, and add them to the optimal button coordinate sequence.

5. The clothing buttonhole precision processing device according to claim 1, characterized in that: The mechanical buttoning operation module comprises: a size error acquisition unit, configured to mechanically button the target fabric according to the optimal button coordinate sequence, and during the buttoning operation, monitor and collect buttoning images of the target fabric, perform buttoning position error analysis based on the buttoning images, acquire transverse size errors and longitudinal size errors, and continuously monitor and generate transverse size error sequences and longitudinal size error sequences; A lateral dimension error prediction unit, used to predict the lateral dimension errors at future K processing points according to the lateral dimension error sequence, and obtain K lateral prediction errors, wherein K is an integer greater than 1; A longitudinal dimension error prediction unit, used to predict longitudinal dimension errors at future K processing points according to the longitudinal dimension error sequence, and obtain K longitudinal prediction errors; A position adaptive correction unit is used to perform position adaptive correction on the target cloth based on the K lateral prediction errors and the K longitudinal prediction errors.

6. The clothing buttonhole precision processing device according to claim 5, characterized in that: The mechanical buttoning operation module comprises: Collect a set of sample lateral error sequences at Q consecutive processing points in the historical processing records, and obtain the lateral dimension errors at subsequent K historical processing points of different sample lateral error sequences to obtain a set of sample lateral prediction error sequences; A sample data equal value division unit is used to divide the sample horizontal error sequence set and the sample horizontal prediction error sequence set into N parts as sample data, select them with replacement N times to construct a first training set, and iterate and select them N times to obtain N training sets; A lateral error prediction channel forming unit is used to respectively train the long short-term memory network using the N training sets until convergence, and obtain N lateral error prediction branches to form a lateral error prediction channel; A branch selection number obtaining unit is used to obtain the number of processing points of the lateral dimension error sequence. If the number of processing points is greater than or equal to Q, the branch selection number is 1. If the number of processing points is less than Q, the ratio of the number of processing points to Q is set to P, and the difference between 1 and P is multiplied by N and rounded to obtain the branch selection number. A lateral error prediction branch selection unit is used to randomly select the lateral error prediction branches of the branch selection number from the N lateral error prediction branches, perform lateral dimension error prediction at future K processing points according to the lateral dimension error sequence, and output the K lateral prediction errors.

7. The precise processing device for buttonholes and buttons of clothing according to claim 5, characterized in that: The mechanical buttoning operation module comprises: A button-stitching operation standard acquisition unit, used to acquire the button-stitching operation standard, wherein the button-stitching operation standard includes a transverse dimension error threshold and a longitudinal dimension error threshold; The earliest lateral warning processing point obtaining unit is used to judge the K lateral prediction errors according to the lateral dimension error threshold, and if there is a situation that exceeds the lateral dimension error threshold, locate the earliest processing point in time to obtain the earliest lateral warning processing point; The earliest longitudinal warning processing point obtaining unit is used to judge the K longitudinal prediction errors according to the longitudinal dimension error threshold, and if there is a situation that exceeds the longitudinal dimension error threshold, locate the earliest processing point in time to obtain the earliest longitudinal warning processing point; The warning processing point determination unit is used to determine the warning processing point according to the earliest horizontal warning processing point and the earliest longitudinal warning processing point, and perform orientation correction on the target cloth according to the initial orientation at the warning processing point.

8. A method for accurately processing buttonholes and buttons on clothing, the method being applied to the accurate processing device for buttonholes and buttons on clothing according to any one of claims 1 to 7, the method comprising: Collect the keyhole images in the target fabric, identify the keyhole image sequence through the convolutional neural network, and obtain the keyhole feature sequence; Performing buttonhole recognition error analysis according to target fabric property information, generating recognition compensation parameters to perform feature compensation on the buttonhole feature sequence, and obtaining a corrected buttonhole feature sequence; Performing iterative optimization analysis of button positions based on the corrected buttonhole feature sequence to generate an optimal button coordinate sequence; According to the optimal button coordinate sequence, the button sewing equipment is controlled to mechanically sew buttons on the target fabric, and the fabric movement trajectory is synchronously monitored during the button sewing operation. The position of the target fabric is adaptively corrected according to the trajectory monitoring result until the operation is completed.

Images