Blended multifunctional fabric pretreatment control method and system based on visual inspection

Abstract

The invention relates to the technical field of visual inspection, and discloses a multifunctional blended fabric pretreatment control method and system based on visual inspection. The method comprises the following steps: acquiring original image data of the surface of the blended fabric; carrying out image gray level conversion and fluff edge feature extraction to obtain fluff distribution data; calculating fluff density and form curvature, and determining a temperature correction sequence; heat distribution information is extracted according to the temperature correction sequence, the matching degree of the heat distribution information and the fluff edge features is calculated, scanning parameters of the industrial camera are dynamically adjusted according to the matching degree, the surface of the fabric subjected to singeing treatment is scanned, and singeing uniformity is obtained; according to the singeing uniformity, a control signal code is generated and transmitted to a singeing device controller, and real-time control is executed after the stability is verified; and optimizing the temperature sequence according to the singeing effect data, and determining the surface uniformity. The uniformity and the stability of the surface quality of the fabric are improved.

Description

Technical Field

[0001] This invention relates to the field of visual inspection technology, and in particular to a method and system for controlling the pretreatment of blended multifunctional fabrics based on visual inspection. Background Technology

[0002] Currently, in the textile industry, singeing is a key pretreatment step for improving the surface smoothness and comfort of blended multifunctional fabrics. Its core lies in the ability to perceive the fabric surface condition online, converting the spatial distribution characteristics of the pile, such as density and height, into precise control commands for the singeing temperature in real time. Therefore, there is an urgent need to introduce industrial vision technology with perception and decision-making capabilities to achieve an intelligent closed loop in the singeing process, from perception to control, thereby ensuring the uniformity and stability of the singeing effect.

[0003] Traditional techniques first rely on preliminary testing of specific blended fabrics to determine a relatively optimized set of singeing process parameters, including flame intensity, machine speed, and singeing distance. During production, operators input these preset parameters into the control system and start the equipment for continuous singeing. During production, operators primarily rely on periodically taking fabric samples from the production line to assess the uniformity of the singeing effect through visual observation and tactile examination. If abnormalities such as residual fibers or localized burns are found, operators manually adjust parameters such as the flame temperature or production line speed based on their personal experience. The entire control process is based on preset values ​​and relies on discrete manual checks for feedback, heavily depending on fixed parameters and human experience. This results in an inability to effectively guarantee the singeing effect and fails to meet the demands of industrial production.

[0004] In improved existing technologies, such as CN113610849B, a scheme for intelligent control of singeing processes using machine vision is disclosed. However, it lacks real-time online monitoring and response to the surface state of the fabric during operation, and cannot perceive and respond to the dynamic changes in the fabric surface nap in real time, resulting in poor adaptability of the singeing process. Therefore, existing technologies suffer from the problem of uneven fabric surface quality. Summary of the Invention

[0005] This invention provides a visual inspection-based pretreatment control method and system for blended multifunctional fabrics to improve fabric surface quality.

[0006] In a first aspect, to address the aforementioned technical problems, this invention provides a visual inspection-based pretreatment control method for blended multifunctional fabrics, comprising: acquiring raw image data of the blended fabric surface using an industrial camera; performing image grayscale conversion processing on the raw image data and extracting pile edge features to obtain pile distribution data; calculating pile density and morphological curvature based on the pile distribution data, extracting regional morphological evaluation results, and determining a temperature correction sequence; extracting heat distribution information from the temperature correction sequence, calculating the matching degree between the heat distribution information and the pile edge features, dynamically adjusting the scanning parameters of the industrial camera based on the matching degree, scanning the fabric surface after singeing treatment, and obtaining singeing uniformity; generating a control signal encoding based on the singeing uniformity and transmitting it to the singeing device controller, and determining the verification stability of the transmission protocol to obtain a stable control sequence; executing a real-time application based on the stable control sequence to obtain singeing effect data, and optimizing the temperature correction sequence based on the singeing effect data to determine surface uniformity.

[0007] In one optional implementation, the step of performing image grayscale conversion processing based on the original image data and extracting fur edge features to obtain fur distribution data includes: performing image grayscale conversion processing based on the original image data using a grayscale weighted average algorithm to obtain grayscale image data including resolution; when the resolution of the grayscale image data is higher than a preset resolution threshold, adjusting the image resolution based on the grayscale image data using a bilinear interpolation method to obtain an adjusted grayscale image; and extracting fur edge features based on the adjusted grayscale image using a Canny edge detection algorithm to obtain fur distribution data.

[0008] In one optional implementation, the step of calculating the fluff density and morphological curvature based on the fluff distribution data, extracting regional morphological evaluation results, and determining the temperature correction sequence includes: calculating the total number of fluff pixels in sub-regions pre-divided by grid division based on the fluff distribution data to obtain a fluff density distribution map of the sub-regions; evaluating the morphological curvature of each sub-region by calculating the density change gradient between each pixel and its neighboring pixels based on the fluff density distribution map to obtain morphological curvature data; classifying the sub-regions using a K-means clustering algorithm based on the morphological curvature data, and extracting variance analysis indicators from the classification results to generate regional morphological evaluation results; logically generating a temperature sequence based on the regional morphological evaluation results using a linear interpolation algorithm combined with preset temperature mapping rules to determine a temperature requirement sequence; generating a distribution heatmap based on the temperature requirement sequence, and adjusting the singeing device parameters using a PID control algorithm to generate a temperature correction sequence.

[0009] In one optional implementation, the step of extracting heat distribution information from the temperature correction sequence, calculating the matching degree between the heat distribution information and the pile edge features, dynamically adjusting the scanning parameters of the industrial camera based on the matching degree, scanning the surface of the singeed fabric, and obtaining singeing uniformity includes: performing region segmentation processing using a grid segmentation algorithm based on the temperature correction sequence to obtain heat distribution data; calculating the matching degree using a cosine similarity algorithm based on the heat distribution data and the pile edge features, combined with a preset feature vector correspondence, to obtain matching degree data; adjusting the scanning frequency and angle range of the industrial camera based on the matching degree data to obtain fabric surface change data containing pixel grayscale values; and calculating the ratio of the standard deviation to the average value of all pixel grayscale values ​​as the singeing uniformity based on the fabric surface change data.

[0010] In one optional implementation, the step of generating a control signal code based on the singeing uniformity and transmitting it to the singeing device controller, and determining the verification stability of the transmission protocol to obtain a stable control sequence, includes: generating a corresponding control signal code based on the singeing uniformity using a pulse width modulation algorithm to obtain an initial control signal sequence; verifying the integrity of transmitted data using the CRC-32 standard in the cyclic redundancy check algorithm based on the initial control signal sequence to obtain a transmission verification result; when the transmission verification result meets a preset stability condition, performing signal smoothing processing using a moving average algorithm based on the initial control signal sequence to extract a stable control sequence; when the transmission verification result does not meet the preset stability condition, calling the most recently valid stable control sequence as a replacement to obtain a stable control sequence.

[0011] In one optional implementation, the step of executing a real-time application based on the stable control sequence to obtain singeing effect data, and optimizing the temperature correction sequence based on the singeing effect data to determine surface uniformity includes: extracting the final control command from the stable control sequence, combining it with a closed-loop control algorithm and a preset signal encoding format definition, transmitting it to the execution module for real-time application to obtain optimized singeing effect data; optimizing and adjusting the temperature correction sequence based on the optimized singeing effect data using a gradient descent algorithm to obtain an optimized temperature correction sequence; and recalculating the matching degree between the heat distribution information and the fluff edge features and calculating the singeing uniformity based on the optimized temperature correction sequence to obtain surface uniformity.

[0012] Secondly, this invention provides a vision-based detection-based pretreatment control system for blended multifunctional fabrics, comprising: a data acquisition module for acquiring raw image data of the blended fabric surface using an industrial camera; a pile distribution analysis module for performing image grayscale conversion processing based on the raw image data and extracting pile edge features to obtain pile distribution data; a temperature correction analysis module for calculating pile density and morphological curvature based on the pile distribution data, extracting regional morphological evaluation results, and determining a temperature correction sequence; a singeing uniformity analysis module for extracting heat distribution information from the temperature correction sequence, calculating the matching degree between the heat distribution information and the pile edge features, dynamically adjusting the scanning parameters of the industrial camera based on the matching degree, scanning the singeed fabric surface, and obtaining singeing uniformity; a control sequence generation module for generating control signal encoding based on the singeing uniformity and transmitting it to the singeing device controller, and determining the verification stability of the transmission protocol to obtain a stable control sequence; and an output module for executing real-time applications based on the stable control sequence to obtain singeing effect data, and optimizing the temperature correction sequence based on the singeing effect data to determine surface uniformity.

[0013] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the visual detection-based pretreatment control method for blended multifunctional fabrics as described in any one of the above.

[0014] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the visual detection-based pretreatment control method for blended multifunctional fabrics described above.

[0015] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention uses visual detection technology to collect images of the surface of blended fabrics in real time, and combines gray-scale weighted average and Canny edge detection algorithm to accurately extract the fluff distribution features, thereby realizing online perception of the surface state of the fabric. This overcomes the limitations of existing technologies that rely on manual sampling and fixed parameters, increases the real-time monitoring and response of the fabric during the singeing process, and improves the real-time response capability of the singeing process. (2) Based on the data of fluff density distribution and morphological curvature, the present invention generates a temperature correction sequence through K-means clustering and linear interpolation algorithm, and combines PID control to dynamically adjust the parameters of the singeing device, thereby achieving precise adaptive control of the thermal field and avoiding local burns or fluff residue problems. (3) The present invention calculates the matching degree between heat distribution and fluff features by using cosine similarity algorithm, and evaluates the singeing uniformity based on the standard deviation of pixel gray value. Combined with pulse width modulation and cyclic redundancy check, it ensures stable transmission of control signal and improves the reliability and consistency of singeing process. (4) Based on the singeing effect data, the present invention uses the gradient descent algorithm to optimize the temperature correction sequence, forming a closed-loop control mechanism, realizing the continuous self-optimization of the singeing process, and improving the surface uniformity and overall quality stability of the blended fabric. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the pretreatment control method for blended multifunctional fabrics based on visual detection provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of the pretreatment control system for blended multifunctional fabrics based on visual detection provided in the second embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Reference Figure 1 The first embodiment of the present invention provides a pretreatment control method for blended multifunctional fabrics based on visual detection, including the following steps: S11, uses an industrial camera to collect raw image data of the surface of the blended fabric; S12, Based on the original image data, perform image grayscale conversion processing and extract the fur edge features to obtain fur distribution data; S13, Based on the fluff distribution data, calculate the fluff density and morphological curvature, extract the regional morphological evaluation results, and determine the temperature correction sequence; S14, extract heat distribution information from the temperature correction sequence, calculate the matching degree between the heat distribution information and the fluff edge features, dynamically adjust the scanning parameters of the industrial camera according to the matching degree, scan the surface of the fabric after singeing treatment, and obtain the singeing uniformity. S15, based on the singeing uniformity, generate a control signal code and transmit it to the singeing device controller, and determine the verification stability of the transmission protocol to obtain a stable control sequence. S16, execute the real-time application according to the stable control sequence to obtain singeing effect data, and optimize the temperature correction sequence according to the singeing effect data to determine the surface uniformity.

[0019] In step S11, raw image data of the surface of the blended fabric is acquired using an industrial camera.

[0020] Specifically, high-resolution industrial cameras deployed at the feed end of the singeing production line continuously capture digital images of the blended fabric surface under uniformly distributed LED lighting conditions. The raw image data is recorded in RGB color mode, containing color information, texture details, and the initial distribution of the pile on the fabric surface. During image acquisition, the shooting frequency of the industrial camera is synchronized with the fabric transport speed to ensure that each segment of the fabric surface is completely captured, avoiding any omissions. This step enables real-time online monitoring of the blended fabric surface condition, providing a raw data foundation for subsequent pile feature extraction and singeing control, overcoming the limitations of traditional methods that rely on manual sampling.

[0021] It should be noted that the frame rate of the industrial camera is set according to the fabric transport speed to ensure sufficient overlap between adjacent images. For example, when the fabric transport speed is 1m / s, the frame rate is no less than 30fps, ensuring that the fabric length covered by each frame is no less than 33mm and the overlap rate between adjacent images is no less than 20%.

[0022] In step S12, the image grayscale conversion is performed based on the original image data, and the fur edge features are extracted to obtain fur distribution data.

[0023] In one specific implementation, the step of performing image grayscale conversion processing based on the original image data and extracting fur edge features to obtain fur distribution data includes: performing image grayscale conversion processing based on the original image data using a grayscale weighted average algorithm to obtain grayscale image data including resolution; when the resolution of the grayscale image data is higher than a preset resolution threshold, adjusting the image resolution based on the grayscale image data using a bilinear interpolation method to obtain an adjusted grayscale image; and extracting fur edge features based on the adjusted grayscale image using a Canny edge detection algorithm to obtain fur distribution data.

[0024] Specifically, in step S12, the original image data is converted to grayscale, and the edge features of the fluff are extracted to obtain fluff distribution data. Specifically, this process first performs grayscale processing on the acquired original image data. The original image data, as input, is an RGB image containing color information from three channels: red, green, and blue. A grayscale weighted average algorithm is used for processing. This algorithm assigns specific weight coefficients to the red, green, and blue channel components (e.g., red component weight is 0.299, green component weight is 0.587, and blue component weight is 0.114). By calculating the sum of the products of the three channel component values ​​of each pixel and their corresponding weight coefficients, the grayscale value of each pixel is obtained, and finally, a grayscale image containing only brightness information is output. The significance of grayscale processing is to simplify the color image into a grayscale image, reducing the amount of data for subsequent calculations, while highlighting the contrast between light and dark areas on the fabric surface, facilitating the identification of fluff features.

[0025] Subsequently, the appropriateness of the resolution of the obtained grayscale image data is assessed. The preset resolution threshold is determined as follows: First, a series of candidate resolutions are selected based on the performance range of the chosen industrial camera. Then, on a fixed hardware platform, the complete image processing workflow is run at different candidate resolutions for various typical blended fabrics, and the processing time and the accuracy of the extracted pile edge features for each frame are recorded. Finally, to meet the minimum real-time frame rate of the system as a processing speed threshold, the highest resolution that ensures the accuracy of pile feature extraction is selected above this threshold, and the number of pixels corresponding to this resolution is set as the resolution threshold (e.g., 1 million pixels). This method ensures that the threshold is both compatible with the camera hardware capabilities and closely related to the actual accuracy of fabric processing and the real-time requirements of the system. The total number of pixels in the grayscale image data is compared with this resolution threshold. If the resolution of the grayscale image data is higher than the resolution threshold, bilinear interpolation is used to downsample the image to adjust the resolution. The implementation process of this method is as follows: For a target pixel, its grayscale value is calculated as the weighted sum of the grayscale values ​​of its four neighboring pixels (Q11, Q12, Q21, Q22) based on the horizontal distance dx and vertical distance dy between the pixel and these four neighboring pixels. The weights are (1-dx)(1-dy), (1-dx)dy, dx(1-dy), and dx*dy, respectively, thereby generating a smoothed grayscale image. This adjustment ensures the processing efficiency of subsequent edge detection algorithms and avoids affecting the real-time performance of the system due to excessive data volume.

[0026] Finally, the edge features of the fuzz are extracted based on the adjusted grayscale image. The Canny edge detection algorithm is employed. This process first uses a Gaussian filter template to convolve the adjusted grayscale image to smooth it and suppress noise. Next, the brightness gradient values ​​of each pixel in the image are calculated in the horizontal and vertical directions, thus obtaining the gradient magnitude and direction. Then, non-maximum suppression is applied to the gradient magnitude, retaining only the points with the largest local magnitude in the gradient direction, forming refined edges. Afterwards, dual thresholds (high and low thresholds) are set for edge connection. The high and low thresholds are determined statistically based on the gradient histogram distribution of a large number of sample images. Points with gradient magnitudes higher than the high threshold are identified as strong edges, points lower than the low threshold are discarded, and points in between are retained as edges if connected to strong edges. Finally, the algorithm outputs a binary image data set, where white pixels represent the detected fuzz edge positions; this data represents the fuzz distribution data. This step accurately identifies and locates the fuzz contours on the fabric surface, transforming visual information into quantifiable spatial distribution data, laying the foundation for subsequent analysis of fuzz density and morphology.

[0027] In step S13, the fiber density and morphological curvature are calculated based on the fiber distribution data, and the regional morphological evaluation results are extracted to determine the temperature correction sequence.

[0028] In one specific implementation, the step of calculating the fluff density and morphological curvature based on the fluff distribution data, extracting regional morphological evaluation results, and determining the temperature correction sequence includes: calculating the total number of fluff pixels in sub-regions pre-divided by grid division based on the fluff distribution data to obtain a fluff density distribution map of the sub-regions; evaluating the morphological curvature of each sub-region by calculating the density change gradient between each pixel and its neighboring pixels based on the fluff density distribution map to obtain morphological curvature data; classifying the sub-regions using a K-means clustering algorithm based on the morphological curvature data, and extracting variance analysis indicators from the classification results to generate regional morphological evaluation results; logically generating a temperature sequence based on the regional morphological evaluation results using a linear interpolation algorithm combined with preset temperature mapping rules to determine a temperature requirement sequence; generating a distribution heatmap based on the temperature requirement sequence, and adjusting the parameters of the singeing device using a PID control algorithm to generate a temperature correction sequence.

[0029] Specifically, in step S13, based on the fluff distribution data, the fluff density and morphological curvature are calculated, and the regional morphological evaluation results are extracted to determine the temperature correction sequence. Specifically, this process first uses the fluff distribution data from step S12 (i.e., a binary image matrix, where white pixels represent fluff) as input, and for pre-divided grid sub-regions, the total number of white pixels in each sub-region is counted. The grid sub-region division method is as follows: First, through a thermal field calibration experiment of the singeing device, the minimum effective area for effective and uniform burning of the heat source on the fabric surface is determined. This area is typically circular or rectangular. Subsequently, using the size of this minimum effective area as a benchmark, the entire fabric surface image is divided into multiple non-overlapping regular grids of the same size, ensuring that each grid sub-region can be independently and precisely controlled by the singeing device. For example, if the calibration shows that the minimum effective range of the thermal field is a square 10 mm wide and 10 mm long, then the image is divided into grids of 10 mm by 10 mm units. This value represents the fluff density of that sub-region, and the density values ​​of all sub-regions together form the fluff density distribution map.

[0030] It should be noted that the grid size is determined according to the thermal field characteristics of the singeing device, and is usually 10mm×10mm to 20mm×20mm; for example, a 10mm×10mm grid can be used for a precision singeing device, and a 20mm×20mm grid can be used for a conventional device.

[0031] Next, based on the villous density distribution map, the density gradient at each pixel's location is obtained by calculating the density difference between each pixel and its eight neighboring pixels. This gradient is a vector containing two components: a horizontal component and a vertical component. Then, the divergence of this gradient vector is calculated to assess the degree of morphological curvature in the vicinity of that point. The specific calculation process for divergence is as follows: First, the differences between the horizontal gradient component of the pixel and the horizontal gradient components of its left and right neighboring pixels, and the differences between the vertical gradient component and the vertical gradient components of its upper and lower neighboring pixels are calculated. Then, these two differences are added together, and the result is the divergence value at that pixel's location. This value quantifies the degree of "convergence" or "divergence" of the gradient field at that point: a positive divergence value indicates that the density gradient around the point is divergent, corresponding to a convex shape similar to a "peak" or "ridge" in the villous distribution; a negative divergence value indicates that the gradient is convergent, corresponding to a concave shape similar to a "valley" or "groove"; the absolute value of the divergence value reflects the severity of morphological curvature. The degree of morphological curvature is evaluated by calculating the Laplacian operator (i.e., the divergence of the gradient) of the density distribution. The absolute value of this operator reflects the severity of density changes and can be used to characterize the local undulations of the villous distribution.

[0032] Then, using the morphological curvature data as input, the K-means clustering algorithm is employed for sub-region classification. The number of clusters in this algorithm, i.e., the final number of morphological categories to be divided, is determined in advance through the elbow rule. The specific implementation steps are as follows: First, a representative historical dataset of blended fabric morphological curvature is collected; then, the number of clusters is incremented from the minimum to the maximum value, and the dataset is clustered multiple times, calculating the sum of squared distances from all sample points to the center point of their respective categories in each clustering result; next, a curve showing the relationship between the number of clusters and the sum of squared distances is plotted. This curve typically changes from a rapid decrease to a gradual decrease as the number of clusters increases. The inflection point of the curve, i.e., the critical point where the effect of reducing the sum of squared distances begins to become insignificant, is selected as the optimal number of clusters, for example, three.

[0033] It should be noted that during the formal clustering process, the algorithm first randomly selects the same number of data points as the preset number of clusters from the morphological curvature data of all sub-regions as initial cluster centers. Then, iterative calculations are performed, each step including two stages: first, each sub-region is assigned to the nearest cluster center based on Euclidean distance; then, the average curvature data of the sub-regions contained in each category is recalculated as the new cluster center. The iteration stops when the position of the cluster center no longer changes significantly or when the preset maximum number of iterations is reached, and the algorithm outputs the final sub-region classification results (e.g., high curvature region, medium curvature region, low curvature region). After clustering, the variance of the curvature values ​​of the sub-regions within each category is calculated, serving as a variance analysis index to generate regional morphological evaluation results that quantify the morphological differences between regions. Subsequently, based on the regional morphological evaluation results and combined with preset temperature mapping rules (which are established based on historical process data statistics, for example, mapping high curvature variance regions to high singeing temperatures), a linear interpolation algorithm is used to smoothly transition the temperature requirements at the boundaries of different category regions.

[0034] The method for establishing the temperature mapping rules is as follows: First, during the system initialization phase, a representative historical production dataset covering various blended fabrics is collected. This dataset contains multiple sets of process parameters corresponding to the optimal singeing effect as determined by expert experience. Each set of parameters includes the fabric region morphology evaluation results before singeing (i.e., region classification, such as high, medium, and low curvature variance regions) and the final applied singeing temperature value. Next, the dataset is cleaned and classified, grouping all samples under the same region morphology classification and their corresponding optimal temperature values. Then, for each region morphology classification, a baseline temperature value is determined through statistical analysis methods (such as calculating the mode of its corresponding temperature value or a densely distributed area within a small range). Finally, a clear rule mapping table is formed, mapping different region morphology categories to their specific baseline temperature values. For example, high curvature variance regions are mapped to a higher baseline singeing temperature, and low curvature variance regions are mapped to a lower baseline singeing temperature. This rule table serves as the preset temperature mapping rule, guiding the generation of temperature sequences in subsequent production. A continuous temperature requirement sequence corresponding one-to-one with each grid sub-region is generated.

[0035] It should be noted that the temperature mapping rule maps the region morphology category to the corresponding reference singeing temperature, which is typically between 600°C and 850°C. For example, low-density, low-curvature regions correspond to 600°C-650°C, medium-density, medium-curvature regions correspond to 700°C-750°C, and high-density, high-curvature regions correspond to 800°C-850°C.

[0036] Finally, the temperature demand sequence is converted into a two-dimensional heat map and used as a setpoint input to the PID control algorithm. This algorithm calculates the error between the temperature demand sequence and the current temperature feedback from the singeing device in real time. Based on the combined calculation results of the proportional, integral, and derivative terms, it dynamically adjusts the power control parameters of the heating element in the singeing device, ultimately outputting a temperature correction sequence that accurately adapts to the current wool density and morphological characteristics. This step transforms the spatial distribution characteristics of the wool into a temperature control command with spatial resolution, achieving localized and refined control of the singeing thermal field. This provides a decision-making basis for overcoming uneven burning or residue problems caused by uneven wool distribution. The parameters of the PID control algorithm are tuned according to the thermal inertia characteristics of the singeing device: the proportional coefficient Kp ranges from 0.5 to 2.0, the integral time Ti ranges from 10 to 30 seconds, and the derivative time Td ranges from 1 to 5 seconds. Specific parameters are determined through engineering tuning methods.

[0037] In step S14, heat distribution information is extracted from the temperature correction sequence, the matching degree between the heat distribution information and the fluff edge features is calculated, the scanning parameters of the industrial camera are dynamically adjusted according to the matching degree, and the surface of the fabric after singeing is scanned to obtain the singeing uniformity.

[0038] In one specific implementation, the step of extracting heat distribution information from the temperature correction sequence, calculating the matching degree between the heat distribution information and the pile edge features, dynamically adjusting the scanning parameters of the industrial camera based on the matching degree, scanning the surface of the singeed fabric, and obtaining singeing uniformity includes: performing region segmentation processing using a grid segmentation algorithm based on the temperature correction sequence to obtain heat distribution data; calculating the matching degree using a cosine similarity algorithm based on the heat distribution data and the pile edge features, combined with a preset feature vector correspondence, to obtain matching degree data; adjusting the scanning frequency and angle range of the industrial camera based on the matching degree data to obtain fabric surface change data containing pixel grayscale values; and calculating the ratio of the standard deviation to the average value of all pixel grayscale values ​​as the singeing uniformity based on the fabric surface change data.

[0039] Specifically, in step S14, heat distribution information is extracted from the temperature correction sequence, the matching degree between the heat distribution information and the pile edge features is calculated, and the scanning parameters of the industrial camera are dynamically adjusted according to the matching degree to scan the surface of the singeed fabric and obtain the singeing uniformity. Specifically, this process uses the temperature correction sequence generated in step S13 as input, which contains the target temperature value for each grid sub-region. Through a grid segmentation algorithm, adjacent sub-regions with the same or similar temperature values ​​are merged to form continuous temperature blocks, thereby obtaining heat distribution data reflecting the thermal energy planning of the entire fabric surface. Next, the matching degree between the heat distribution data and the pile edge feature data from step S12 is calculated: First, a feature vector is constructed for each grid region according to a preset feature vector correspondence. The preset method for this correspondence is as follows: during the system deployment phase, key physical quantities for matching are determined based on a large amount of historical data.

[0040] Specifically, the "regional heat distribution intensity value" (i.e., the target temperature value of the region in the temperature correction sequence), which characterizes the application of heat energy, is used as the first vector element, and the "regional fluff edge density value" (i.e., the total number of edge pixels obtained from fluff distribution data within the region), which characterizes the distribution of fluff, is used as the second vector element. Thus, a feature vector containing these two dimensions is constructed for each region, quantifying the region's state from both heat energy and fluff perspectives. This rule of mapping the physical state of a region to a fixed-dimensional vector constitutes the predefined feature vector correspondence. Subsequently, a cosine similarity algorithm is applied to calculate the cosine of the angle between these two feature vectors in multidimensional space. This value is obtained by dividing the dot product of the two vectors by their respective magnitudes (i.e., the product of vector lengths). The result serves as the matching degree data; the closer the value is to 1, the more ideal the spatial correspondence between the heat distribution and the fluff distribution.

[0041] Based on the calculated matching degree data, the scanning parameters of the industrial camera used to detect the singeing effect are dynamically adjusted. If the matching degree is high, the scanning frequency and angle range are maintained or fine-tuned. If the matching degree is low, the scanning frequency is increased or the angle range is expanded according to the adjustment rules determined based on historical experimental data to obtain more detailed surface data. For example, if the matching degree data is lower than a preset matching degree threshold (e.g., 0.85), the scanning frequency of the industrial camera is increased (e.g., from 30fps to 45fps) or its horizontal scanning angle range is expanded (e.g., from ±15° to ±25°) according to a preset adjustment scale to obtain more detailed surface data. The matching degree threshold and adjustment scale are determined by testing different fabric samples in the early stage to ensure that effective image data can be re-acquired. After adjustment, the industrial camera scans the surface of the singeed fabric to obtain fabric surface change data reflecting the change in the degree of surface ablation. This data consists of a grayscale matrix of each pixel.

[0042] Finally, based on the fabric surface change data, the singeing uniformity is evaluated by calculating the arithmetic mean of the grayscale values ​​of all pixels within the entire detection area, along with the standard deviation of these grayscale values. The singeing uniformity is ultimately quantified as the ratio of the standard deviation to the mean; a smaller ratio indicates better grayscale consistency and higher uniformity on the fabric surface, i.e., higher singeing uniformity. This step, by quantifying the matching accuracy between heat application and pile distribution, as well as the final singeing effect uniformity, provides a direct and quantitative evaluation basis for judging the effectiveness of the control strategy and subsequent temperature sequence optimization.

[0043] In step S15, based on the singeing uniformity, a control signal is generated and transmitted to the singeing device controller, and the stability of the transmission protocol is checked to obtain a stable control sequence.

[0044] In one specific implementation, the step of generating a control signal code based on the singeing uniformity and transmitting it to the singeing device controller, and determining the verification stability of the transmission protocol to obtain a stable control sequence, includes: generating a corresponding control signal code based on the singeing uniformity using a pulse width modulation algorithm to obtain an initial control signal sequence; verifying the integrity of transmitted data using the CRC-32 standard in the cyclic redundancy check algorithm based on the initial control signal sequence to obtain a transmission verification result; when the transmission verification result meets a preset stability condition, performing signal smoothing processing using a moving average algorithm based on the initial control signal sequence to extract a stable control sequence; when the transmission verification result does not meet the preset stability condition, calling the most recently valid stable control sequence as a replacement to obtain a stable control sequence.

[0045] Specifically, this process uses the singeing uniformity value calculated in step S14 as input. This singeing uniformity value is a key indicator for evaluating the deviation between the current singeing effect and the ideal state. It should be noted that the control object here is not the overall power of the singeing device, but rather the temperature correction sequence generated in step S13 for each grid sub-region on the fabric surface. This sequence itself already contains a spatial temperature distribution that matches the pile distribution, pre-set according to the pile density and morphology. The specific control object of the control sequence is the array of independent heating units in the singeing device that corresponds one-to-one with the grid sub-regions, such as an infrared heating tube array or multiple independent control burners of a gas burner. The role of the singeing uniformity value here is not to directly generate a new control sequence, but to serve as a feedback signal for overall optimization and fine-tuning of the current set of temperature control commands (i.e., temperature correction sequence) that already possess spatial distribution characteristics.

[0046] The specific adjustment logic is as follows: The system presets a baseline singeing uniformity threshold based on historical production data statistics. This threshold is calculated during the system initialization phase by analyzing the singeing uniformity data corresponding to a large number of qualified products, taking their statistical average, and adding three times the standard deviation to ensure that it represents an excellent level under stable processes. When the singeing uniformity calculated this time deviates from this baseline threshold, the system will initiate overall optimization of the temperature correction sequence. This optimization process is based on a linear scaling rule determined by fitting a large amount of process experimental data in the early stages. The core of this rule is to establish a continuous mapping relationship from the singeing uniformity deviation value to the temperature scaling ratio.

[0047] Specifically, the relative deviation ratio between the current singeing uniformity and the baseline threshold is first calculated. Then, this deviation ratio is multiplied by a fixed scaling factor determined through regression analysis of historical data to obtain an overall temperature adjustment ratio. Finally, this adjustment ratio is applied to each temperature value in the temperature correction sequence, achieving proportional and directional scaling of the temperature across all sub-regions. This scaling process maintains the spatial distribution relationship defined by the original temperature correction sequence, preserving the relative relationship of "higher temperature where there is more singeing and lower temperature where there is less singeing," and only performs global calibration on the overall intensity level of the thermal field.

[0048] Subsequently, based on this optimized and fine-tuned final temperature correction sequence, a corresponding control signal code, i.e., the initial control signal sequence, is generated using a pulse width modulation algorithm. Each pulse signal in this sequence corresponds to an independent heating unit, and its pulse width is proportional to the final set temperature value of the sub-region where the heating unit is located. In this way, the control signal code accurately converts the temperature control command, which includes spatial distribution information and overall intensity calibration, to achieve refined closed-loop control of the thermal field of the singeing device.

[0049] It should be noted that, to ensure the accuracy of the instructions transmitted to the singeing unit controller, the initial control signal sequence undergoes data integrity verification. This is done using the CRC-32 standard of the Cyclic Redundancy Check algorithm: the initial control signal sequence is treated as a long binary number. This binary number is divided by a fixed, standard 32-power generator polynomial (whose coefficients are defined by the CRC-32 standard). The resulting 32-bit remainder is the checksum, which is appended to the initial control signal sequence to form the transmitted data packet. The receiving controller performs the same calculation. If the remainder matches the received checksum, the transmission verification result is marked as "valid," indicating that no bit errors occurred during data transmission; otherwise, it is marked as "invalid," indicating that the data transmission is unreliable. The preset stability condition is that the transmission verification result must be "valid." This is a binary judgment condition with no intermediate states or thresholds. Its core requirement is that the checksum must match perfectly to ensure the absolute accuracy of the control instructions. In specific application scenarios, an acceptable bit error rate threshold can be set according to the reliability of the communication protocol, but the core principle is to ensure the accuracy and execution security of the control instructions.

[0050] If the transmission verification result meets the stability condition, the initial control signal sequence is further smoothed using a moving average algorithm. This algorithm sets a time window size based on the controller response cycle, calculates the arithmetic average of the initial control signal values ​​at the current moment and several moments prior within the window, and replaces the current signal value with the average value, thereby filtering out possible random fluctuations or glitches in the signal and extracting a stable control sequence with gradual changes. If the transmission verification result does not meet the stability condition, it indicates that the transmitted data may be erroneous. In this case, the system automatically calls the most recently verified valid stable control sequence from the storage unit as the alternative output for this control, thereby ensuring that the singeing device can continue to operate based on a known valid control command and maintain the continuity of the process. This step reliably transforms the singeing uniformity evaluation result into an anti-interference and stable execution command, ensuring the accuracy of control signal transmission and the stability of the execution process.

[0051] In step S16, a real-time application is executed according to the stable control sequence to obtain singeing effect data, and the temperature correction sequence is optimized based on the singeing effect data to determine the surface uniformity.

[0052] In one specific implementation, the step of executing a real-time application based on the stable control sequence to obtain singeing effect data, and optimizing the temperature correction sequence based on the singeing effect data to determine surface uniformity includes: extracting the final control command from the stable control sequence, combining it with a closed-loop control algorithm and a preset signal encoding format definition, transmitting it to the execution module for real-time application to obtain optimized singeing effect data; optimizing and adjusting the temperature correction sequence based on the optimized singeing effect data using a gradient descent algorithm to obtain an optimized temperature correction sequence; and recalculating the matching degree between the heat distribution information and the fluff edge features and calculating the singeing uniformity based on the optimized temperature correction sequence to obtain surface uniformity.

[0053] Specifically, in step S16, real-time application is executed according to the stable control sequence to obtain singeing effect data, and the temperature correction sequence is optimized based on the singeing effect data to determine surface uniformity. Specifically, this process first uses the stable control sequence generated in step S15 as the final control command, which is a pulse width modulation signal sequence after verification and smoothing. According to a predefined signal encoding format, this command is transmitted to the heating element execution module of the singeing device through a closed-loop control loop, driving it to precisely adjust the heat source power and action time according to the command, thereby achieving real-time singeing treatment of the fabric surface. During this process, images of the processed fabric are synchronously acquired through an online visual sensor, and the pixel grayscale value matrix obtained after grayscale processing is the optimized singeing effect data reflecting the effect of this singeing treatment.

[0054] Subsequently, using the optimized singeing effect data as an evaluation basis, the temperature correction sequence generated in step S13 is optimized. A gradient descent algorithm is employed, treating the temperature value of each sub-region in the temperature correction sequence as an adjustable parameter, and using the overall singeing uniformity (i.e., the ratio of the standard deviation to the mean of grayscale values) calculated from the optimized singeing effect data as the objective function. The algorithm determines the parameter adjustment direction and step size to improve the singeing uniformity towards the optimal value (minimum ratio) by calculating the partial derivative (i.e., gradient) of the objective function with respect to each temperature parameter. The step size is controlled by the learning rate parameter determined through historical data experiments. Through iterative calculation, the temperature values ​​of each sub-region are continuously fine-tuned until the change in the objective function is less than the preset convergence threshold or the maximum number of iterations is reached, ultimately generating a new set of optimized temperature correction sequences that theoretically produce better singeing effects.

[0055] Finally, the optimized temperature correction sequence is used as new input to re-execute the complete process of step S14. However, in this optimization loop, the fluff edge feature data used for matching needs to be updated synchronously: First, the surface image of the blended fabric after the current round of singeing is acquired again using an industrial camera, and the new image is converted to grayscale and edge features are extracted using the same method described in step S12 to obtain fluff distribution data reflecting the latest surface state. Subsequently, new thermal distribution data is generated based on the optimized temperature correction sequence, and its matching degree with the newly extracted fluff edge feature data is calculated. Based on this matching degree, the sensor parameters are adjusted, and the fabric surface change data is acquired again. Finally, a new singeing uniformity value is calculated, which is determined as the surface uniformity characterizing the final processing effect. This step constitutes a complete, self-optimizing closed-loop control system that can automatically adjust the control strategy according to the feedback of each processing effect, so that the temperature parameter setting continuously approaches the optimal solution, thereby continuously improving the uniformity and stability of the surface quality of the blended fabric after singeing.

[0056] Reference Figure 2 The second embodiment of the present invention provides a vision-based detection-based pretreatment control system for blended multifunctional fabrics, comprising: a data acquisition module for acquiring raw image data of the blended fabric surface using an industrial camera; a pile distribution analysis module for performing image grayscale conversion processing based on the raw image data and extracting pile edge features to obtain pile distribution data; a temperature correction analysis module for calculating pile density and morphological curvature based on the pile distribution data, extracting regional morphological evaluation results, and determining a temperature correction sequence; a singeing uniformity analysis module for extracting heat distribution information from the temperature correction sequence, calculating the matching degree between the heat distribution information and the pile edge features, dynamically adjusting the scanning parameters of the industrial camera based on the matching degree, scanning the singeed fabric surface, and obtaining singeing uniformity; a control sequence generation module for generating control signal encoding based on the singeing uniformity and transmitting it to the singeing device controller, and judging the verification stability of the transmission protocol to obtain a stable control sequence; and an output module for executing real-time applications based on the stable control sequence to obtain singeing effect data, and optimizing the temperature correction sequence based on the singeing effect data to determine surface uniformity.

[0057] It should be noted that the visual detection-based pretreatment control device for blended multifunctional fabrics provided in this embodiment of the invention is used to execute all the process steps of the visual detection-based pretreatment control method for blended multifunctional fabrics described in the above embodiment. The working principles and beneficial effects of the two are one-to-one, and therefore will not be described in detail again.

[0058] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a vision-detection-based pretreatment control program for blended multifunctional fabrics. When the processor executes the computer program, it implements the steps in the various vision-detection-based pretreatment control method embodiments for blended multifunctional fabrics described above, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as a vision-based detection-based pretreatment control module for blended multifunctional fabrics.

[0059] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0060] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0061] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0062] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0063] If the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0064] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0065] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for controlling the pretreatment of blended multifunctional fabrics based on visual inspection, characterized in that, include: Raw image data of the blended fabric surface is acquired using an industrial camera. Based on this raw image data, grayscale conversion is performed, and pile edge features are extracted to obtain pile distribution data. Based on the pile distribution data, pile density and morphological curvature are calculated, and regional morphological evaluation results are extracted to determine a temperature correction sequence. Thermal distribution information is extracted from the temperature correction sequence, and the matching degree between the thermal distribution information and the pile edge features is calculated. The scanning parameters of the industrial camera are dynamically adjusted based on the matching degree to scan the singeed fabric surface and obtain singe uniformity. Based on the singe uniformity, a control signal is generated and encoded, transmitted to the singeing device controller, and the stability of the transmission protocol is checked to obtain a stable control sequence. Real-time application is executed based on the stable control sequence to obtain singe effect data. The temperature correction sequence is then optimized based on the singe effect data to determine surface uniformity.

2. The pretreatment control method for blended multifunctional fabrics based on visual inspection according to claim 1, characterized in that, The step of performing image grayscale conversion processing based on the original image data and extracting fur edge features to obtain fur distribution data includes: performing image grayscale conversion processing based on the original image data using a grayscale weighted average algorithm to obtain grayscale image data including resolution; when the resolution of the grayscale image data is higher than a preset resolution threshold, adjusting the image resolution based on the grayscale image data using a bilinear interpolation method to obtain an adjusted grayscale image; and extracting fur edge features based on the adjusted grayscale image using a Canny edge detection algorithm to obtain fur distribution data.

3. The pretreatment control method for blended multifunctional fabrics based on visual inspection according to claim 1, characterized in that, The step of calculating the fluff density and morphological curvature based on the fluff distribution data, extracting regional morphological evaluation results, and determining the temperature correction sequence includes: calculating the total number of fluff pixels in sub-regions pre-divided by gridding based on the fluff distribution data to obtain a fluff density distribution map for each sub-region; evaluating the morphological curvature of each sub-region by calculating the density change gradient between each pixel and its neighboring pixels based on the fluff density distribution map to obtain morphological curvature data; classifying sub-regions using a K-means clustering algorithm based on the morphological curvature data, and extracting variance analysis indicators from the classification results to generate regional morphological evaluation results; logically generating a temperature sequence based on the regional morphological evaluation results using a linear interpolation algorithm combined with preset temperature mapping rules to determine the temperature requirement sequence; generating a distribution heatmap based on the temperature requirement sequence, and adjusting the parameters of the singeing device using a PID control algorithm to generate the temperature correction sequence.

4. The pretreatment control method for blended multifunctional fabrics based on visual inspection according to claim 1, characterized in that, The process of extracting heat distribution information from the temperature correction sequence, calculating the matching degree between the heat distribution information and the pile edge features, dynamically adjusting the scanning parameters of the industrial camera based on the matching degree, scanning the surface of the singeed fabric, and obtaining singeing uniformity includes: performing region segmentation processing using a grid segmentation algorithm based on the temperature correction sequence to obtain heat distribution data; calculating the matching degree using a cosine similarity algorithm based on the heat distribution data and the pile edge features, combined with a preset feature vector correspondence, to obtain matching degree data; adjusting the scanning frequency and angle range of the industrial camera based on the matching degree data to obtain fabric surface change data containing pixel grayscale values; and calculating the ratio of the standard deviation to the average value of all pixel grayscale values ​​as the singeing uniformity based on the fabric surface change data.

5. The pretreatment control method for blended multifunctional fabrics based on visual inspection according to claim 1, characterized in that, The process of generating a control signal code based on the singeing uniformity and transmitting it to the singeing device controller, and determining the verification stability of the transmission protocol to obtain a stable control sequence, includes: generating a corresponding control signal code based on the singeing uniformity using a pulse width modulation algorithm to obtain an initial control signal sequence; verifying the integrity of transmitted data using the CRC-32 standard in the cyclic redundancy check algorithm based on the initial control signal sequence to obtain a transmission verification result; when the transmission verification result meets a preset stability condition, performing signal smoothing processing using a moving average algorithm based on the initial control signal sequence to extract a stable control sequence; and when the transmission verification result does not meet the preset stability condition, using the most recently valid stable control sequence as a replacement to obtain a stable control sequence.

6. The pretreatment control method for blended multifunctional fabrics based on visual detection according to claim 1, characterized in that, The process of executing a real-time application based on the stable control sequence to obtain singeing effect data, and optimizing the temperature correction sequence based on the singeing effect data to determine surface uniformity includes: extracting the final control command from the stable control sequence, combining it with a closed-loop control algorithm and a preset signal encoding format definition, transmitting it to the execution module for real-time application to obtain optimized singeing effect data; optimizing and adjusting the temperature correction sequence based on the optimized singeing effect data using a gradient descent algorithm to obtain an optimized temperature correction sequence; recalculating the matching degree between heat distribution information and pile edge features based on the optimized temperature correction sequence; dynamically adjusting the scanning parameters of the industrial camera based on the matching degree; scanning the surface of the singeed fabric to obtain surface uniformity.

7. A vision-based detection-based pretreatment control system for blended multifunctional fabrics, characterized in that, include: The data acquisition module is used to acquire raw image data of the surface of the blended fabric using an industrial camera; The fluff distribution analysis module is used to perform image grayscale conversion processing based on the original image data, and extract fluff edge features to obtain fluff distribution data; The temperature correction analysis module is used to calculate the fiber density and morphological curvature based on the fiber distribution data, extract the regional morphological evaluation results, and determine the temperature correction sequence. The singeing uniformity analysis module is used to extract heat distribution information from the temperature correction sequence, calculate the matching degree between the heat distribution information and the fluff edge features, dynamically adjust the scanning parameters of the industrial camera according to the matching degree, scan the fabric surface after singeing treatment, and obtain the singeing uniformity; the control sequence generation module is used to generate control signal encoding based on the singeing uniformity and transmit it to the singeing device controller, and determine the verification stability of the transmission protocol to obtain a stable control sequence. The output module is used to execute a real-time application based on the stable control sequence to obtain singeing effect data, and optimize the temperature correction sequence based on the singeing effect data to determine the surface uniformity.

Images