A method for detecting garment wrinkles based on image analysis

Abstract

The application discloses a kind of clothing wrinkle detection methods based on image analysis, it is related to image data processing technical field, comprising: according to the feature atlas after denoising, preliminary positioning is carried out to wrinkle area, region division is carried out to image data under dynamic scene, and the wrinkle area of preliminary segmentation is obtained;Through the wrinkle area of preliminary segmentation, the boundary of region is optimized and handled, and clear wrinkle boundary data is obtained;According to the detection information of real-time transmission, generate detection report data, integrate information in combination with production line feedback mechanism, obtain comprehensive data for quality evaluation;Through comprehensive data, detection information and production parameter are associated analysis, if detection information shows that wrinkle exceeds tolerance range, then automatically adjust image acquisition parameter, and obtain the detection process data of optimization.The application realizes the accurate detection of clothing wrinkle under complex environment.

Description

Technical Field

[0001] This invention belongs to the field of image data processing technology, and in particular relates to a method for detecting clothing wrinkles based on image analysis. Background Technology

[0002] In assessing the smoothness and aesthetics of garments, wrinkle detection has become a crucial step affecting production efficiency and quality control. However, current mainstream detection methods have significant limitations. Traditional methods rely on manual judgment, which is inefficient and easily influenced by subjective factors. While automation technology has made progress, it often performs poorly in complex environments and fails to meet practical needs.

[0003] Specifically, existing methods often suffer from decreased detection accuracy when dealing with diverse clothing materials and complex fold patterns due to insufficient adaptability to environmental changes. This lack of adaptability stems primarily from limited ability to handle different lighting conditions and shooting angles, making detection results susceptible to external interference. As this problem intensifies, it further exposes the limitations of existing technologies in handling small or dense folds, especially in multi-angle or dynamic scenes, where capturing subtle features becomes a major challenge. This challenge directly leads to another related issue: existing technologies struggle to achieve rapid detection while maintaining high accuracy, particularly on industrial production lines where a balance between real-time performance and accuracy is virtually impossible.

[0004] Therefore, how to improve the adaptability and accuracy of wrinkle detection in complex environments, while taking into account real-time performance and resource efficiency, has become a key issue that this study urgently needs to address. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes an image analysis-based method for detecting clothing folds, which enables accurate detection of clothing folds in complex environments.

[0006] To achieve the above objectives, the present invention provides a clothing wrinkle detection method based on image analysis, comprising: acquiring clothing image data; annotating the clothing image data to obtain an initial training dataset; extracting features from the initial training dataset to obtain a feature map; and filtering the feature map to obtain a denoised feature map.

[0007] The denoised feature map is divided into regions to obtain preliminarily segmented wrinkled regions; the boundaries of the preliminarily segmented wrinkled regions are optimized to obtain clear wrinkle boundary data; the clear wrinkle boundary data is combined with a preset wrinkle shape template for matching analysis to obtain the final wrinkle detection result.

[0008] The final wrinkle detection result is synchronized with the image acquisition frequency of the production line to obtain real-time transmitted detection information; based on the real-time transmitted detection information, detection report data is automatically generated, and information is integrated with the production line feedback mechanism to obtain comprehensive data; the comprehensive data is correlated with production parameters to obtain garment wrinkle detection results.

[0009] Technical Effects of this Invention: This invention discloses a garment wrinkle detection method based on image analysis. It obtains and annotates garment image data under different materials and lighting conditions by pre-establishing a multi-scene image database, thus generating an initial training dataset. A convolutional neural network model is used to extract features and perform multi-scale analysis on the image data, resulting in a feature map containing subtle features. Noise interference caused by lighting variations is filtered out, and an image segmentation algorithm is used to initially locate the wrinkle region. An edge enhancement algorithm is then used to optimize the region boundaries. Finally, a matching analysis is performed using a pre-set wrinkle shape template to obtain the final wrinkle detection result. This invention achieves accurate detection of garment wrinkles in complex environments and can be synchronized with the production line in real time, automatically generating detection reports, providing an effective solution for garment production quality control. Attached Figure Description

[0010] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0011] Figure 1 This is a schematic flowchart of an image analysis-based clothing wrinkle detection method according to an embodiment of the present invention. Detailed Implementation

[0012] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0013] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0014] like Figure 1 As shown, this embodiment provides a clothing wrinkle detection method based on image analysis, including: acquiring clothing image data; annotating the clothing image data to obtain an initial training dataset; extracting features from the initial training dataset to obtain a feature map; and filtering the feature map to obtain a denoised feature map.

[0015] The denoised feature map is divided into regions to obtain preliminarily segmented wrinkled regions; the boundaries of the preliminarily segmented wrinkled regions are optimized to obtain clear wrinkle boundary data; the clear wrinkle boundary data is combined with a preset wrinkle shape template for matching analysis to obtain the final wrinkle detection result.

[0016] The final wrinkle detection result is synchronized with the image acquisition frequency of the production line to obtain real-time transmitted detection information; based on the real-time transmitted detection information, detection report data is automatically generated, and information is integrated with the production line feedback mechanism to obtain comprehensive data; the comprehensive data is correlated with production parameters to obtain garment wrinkle detection results.

[0017] Furthermore, obtaining the initial training dataset includes:

[0018] Acquire clothing image data containing multiple scene images, classify and organize the clothing image data, perform preliminary noise reduction processing on the clothing image data, and obtain a clear first image set;

[0019] Each image in the first image set is processed to distinguish regions, identify wrinkled and non-wrinkled regions in the images, and mark images whose detected edge features meet a preset threshold as wrinkled regions to obtain the initial training dataset.

[0020] Specifically, when constructing a clothing image database, images of clothing in different scenes can be obtained from multiple public datasets or self-built datasets, covering street, indoor, and outdoor environments to ensure data diversity. When classifying and organizing images based on material differences and lighting conditions, images can be categorized into materials such as cotton, silk, and leather, and further divided into three categories based on lighting intensity: strong light, natural light, and low light. The number of images in each category is controlled to around 1000 to ensure the balance of the initial data. In the image denoising stage, image processing tools such as median filtering are used to initially process the clothing images, removing noise points caused by the shooting equipment or environment. When processing a cotton shirt image, noise points caused by light reflection can be observed in the original image. After filtering, the image details are clearer, forming the first image set, providing high-quality input for subsequent region differentiation. For region differentiation operations, edge detection tools such as the Canny operator are used to identify wrinkled and non-wrinkled regions.

[0021] Furthermore, obtaining the feature map includes:

[0022] The clothing image data in the initial training dataset is decomposed into layers, and the decomposed images are divided into regions to obtain a set of layered image data.

[0023] Using the layered image data set, multi-scale features are extracted using an edge detection tool. If the detected feature intensity is lower than a preset threshold, the clothing image data is enhanced to obtain an intermediate feature set containing subtle features.

[0024] Based on the intermediate feature set, the features are integrated, and features at different scales are weighted and combined to obtain a unified feature representation map; the unified feature representation map is used to distinguish image content, and combined with the data feature analysis results, a feature map is obtained.

[0025] Specifically, when processing images based on the initial training dataset, a hierarchical decomposition can be performed on clothing images considering illumination variations and material diversity. For illumination variations, images can be initially divided into strong light, natural light, and weak light. Image processing tools can then be used to analyze the brightness distribution under each type of illumination, obtaining a hierarchical image dataset. Taking cotton clothing images under strong light as an example, their brightness values ​​may be concentrated in a higher range. Hierarchical decomposition can clearly define the feature distribution of bright and shadow areas, providing a clear basis for subsequent processing. When using edge detection tools to extract multi-scale features, different detection parameters can be set for clothing images of different materials. For silk clothing, due to its smooth surface and significant reflective properties, a lower feature intensity threshold, such as 30, can be set to capture subtle texture variations. If the detected feature intensity is below this threshold, enhancement processing is used to improve detail representation, forming an intermediate feature set containing subtle features. This approach ensures that key detail information is preserved even under complex materials.

[0026] Furthermore, the denoised feature maps are obtained as follows:

[0027] Based on the feature map, the image content is divided into regions, the parts affected by noise are separated, and a preliminary set of feature regions is obtained.

[0028] Based on the initially defined set of feature regions, brightness standardization is performed on each region. If the detected grayscale deviation exceeds a preset threshold, brightness correction is performed on the region to determine the corrected set of feature regions.

[0029] Based on the corrected feature region set, noise interference is suppressed while preserving the integrity of subtle features, resulting in a denoised feature map.

[0030] Specifically, when segmenting image content regions based on feature maps, image segmentation tools can be used to address environmental interference in clothing images. For complex background noise, the tool can initially divide the image content into main and background regions, separating potentially affected areas. Taking a cotton shirt image as an example, in an indoor scene, background clutter and shadows may affect the subtle features of the shirt's edges. Segmentation tools can initially identify these affected areas, forming a preliminary set of feature regions, laying the foundation for subsequent processing. When using grayscale adjustment tools to standardize the brightness of the initially segmented feature regions, the impact of lighting changes can be considered. Assuming a preset grayscale deviation threshold of 20, if the grayscale deviation of a certain area reaches 25, brightness correction is required. Taking a silk dress image under strong light as an example, the highlight areas on the dress surface may have higher grayscale values ​​due to uneven lighting. By adjusting the brightness distribution to make it consistent with the grayscale values ​​of areas under natural light, a corrected set of feature regions is formed. This method helps to unify the visual appearance of different areas. When performing noise suppression based on a corrected feature region set, filtering tools can address background noise or lighting interference while preserving the integrity of subtle features. Taking a leather garment image as an example, fine surface cracks are key features. Direct filtering might lead to loss of detail; therefore, gentler filtering parameters can be chosen, such as setting the filter radius to 2 pixels, to ensure that texture details are preserved while denoising, resulting in a denoised feature data set. This approach effectively balances noise suppression and feature preservation.

[0031] Furthermore, obtaining clear fold boundary data includes:

[0032] Based on the denoised feature map, the pixel gradients around the region are analyzed to extract the key parts of the blurred boundary and obtain preliminary boundary information. Using the preliminary boundary information, the pixel values ​​of the boundary part are processed to enhance the contrast of the boundary region and obtain enhanced boundary feature data.

[0033] Based on the enhanced boundary feature data, noise within the boundary area is cleaned up to obtain cleaned boundary lines; the continuity of the boundary lines is detected using the cleaned boundary lines, and interrupted or blurred parts are repaired to obtain clear folded boundary data.

[0034] Specifically, when processing wrinkled areas in clothing images, various techniques can be used to refine the boundaries of initially segmented areas, especially in dynamic scenes where boundary blurring and interference are common. The following analysis and examples focus on each technique within the field of clothing image processing. Edge detection tools analyze the pixel gradients around a region, which can be understood as identifying boundaries by detecting changes in pixel brightness. In dynamic scenes, complex environments can lead to blurred boundaries, such as uneven lighting or clothing movement. Based on the initially identified boundary information, image sharpening tools are applied to enhance the contrast of the boundary areas. Dynamic interference, such as clothing shaking caused by wind, can cause uneven distribution of boundary pixel values. Taking the collar area of ​​a white shirt as an example, in outdoor scenes, changes in background lighting can weaken boundary clarity. Sharpening tools adjust boundary pixel values, such as increasing the contrast to a higher value within a preset range, like from the original value of 10 to 15, highlighting the difference between the boundary and the background, resulting in enhanced feature data. This processing makes the boundaries easier to identify subsequently. For the enhanced boundary feature data, filtering tools are used to clean up noise within the boundary areas. In dynamic scenes, noise may originate from background clutter or lighting interference. Taking the hem area of ​​a shirt as an example, tiny bright spots in the background might be misidentified as boundary areas. If the detected noise value exceeds a preset threshold (e.g., a set value of 5 but an actual value of 7), the filtering tool will perform local smoothing on that area, preserving the boundary lines while removing irrelevant interference to obtain cleaned lines. This method effectively reduces misjudgments.

[0035] Furthermore, obtaining the final wrinkle detection results includes:

[0036] Based on the clear fold boundary data, preliminary extraction of morphological features is performed. Combined with a preset morphological template library, the morphological matching degree of the boundary data is compared to obtain a preliminary comparison result set. The regional lines in the preliminary comparison result set are supplemented to obtain an adjusted boundary dataset.

[0037] Based on the adjusted boundary dataset, the detection accuracy value is checked a second time. If the detection accuracy value is lower than the preset threshold, the boundary data is optimized to determine the optimized boundary feature set.

[0038] The boundary clarity is finally verified using the optimized boundary feature set. The correction data is then integrated using an image fusion tool to obtain the final wrinkle detection result.

[0039] Specifically, based on the preliminary comparison result set, the boundary correction method aims to locally adjust inconsistent marked areas. Its core lies in using algorithms to analyze the continuity of boundary lines and supplementing missing parts with contour repair tools. Taking the sleeve folds of a suit jacket as an example, if the boundary lines break at certain points due to folding, the contour repair tool will automatically supplement the missing parts according to the direction of surrounding lines, for example, connecting lines with a break length of approximately 2 pixels to form an adjusted boundary dataset. This method ensures the integrity of the boundary data. For the adjusted boundary dataset, a secondary check using a threshold judgment method is used to verify the detection accuracy. The principle is to evaluate the boundary data using a preset accuracy threshold, such as 85%. If the detection accuracy is lower than this threshold, it is optimized using a correction iteration method. Taking the waist fold area of ​​a suit jacket as an example, if the initial detection accuracy is only 80%, the correction iteration method will fine-tune the boundary data multiple times based on the features of surrounding pixels until the accuracy reaches or exceeds the threshold, ultimately determining the optimized boundary feature set. This iterative adjustment process helps improve the reliability of the data.

[0040] Furthermore, the real-time transmitted detection information includes:

[0041] The acquired image data and the final wrinkle detection result are synchronously mapped on the time axis to obtain a time-consistent data set after matching; the redundant information in the detection result is simplified through the time-consistent data set, and the simplified data is hierarchically packaged to obtain a simplified data package;

[0042] The simplified data packet is sent in segments, and the data integrity of the sent simplified data packet is checked to determine the complete data unit in the transmission process. If the complete data unit passes the check, the segmented data is recombined, and the recombined detection information is pushed by the information distribution component to obtain the real-time transmission detection information.

[0043] Specifically, in the field of garment image processing, efficient management of image acquisition frequency is a crucial aspect for meeting the real-time processing needs of production lines. The application of timestamp alignment tools aims to ensure the consistency of acquired image data and detection results along the timeline. Taking a white shirt being rapidly scanned on the production line as an example, assuming an image acquisition frequency of 10 frames per second and a detection result generation frequency of 8 times per second, timestamp alignment tools can map each frame of image to the most recent detection result, forming a time-consistent data set. This method effectively avoids data misalignment problems caused by inconsistent frequencies. For time-consistent data sets, data compression tools are used to handle redundant information in the detection results. Taking the wrinkle detection data of the shirt collar area as an example, the original data may contain a large number of repetitive pixel feature descriptions. Data compression tools will simplify these redundant parts; assuming the original data size is 5MB, it may be reduced to 2MB after compression. Considering storage efficiency requirements, hierarchical encoding methods will package data according to importance, for example, setting key collar boundary data as high priority and other secondary area data as low priority, forming a streamlined data package. This hierarchical approach helps optimize storage space allocation.

[0044] Furthermore, obtaining comprehensive data includes:

[0045] Based on the real-time transmitted detection information, key quality-related fields are separated from the information stream, a set of separated quality fields is obtained, and the core data units in the set are determined. The core data units are mapped to a pre-established report format to obtain the mapped formatted data group, and the data content that conforms to the report structure is determined.

[0046] If the formatted data group meets the preset threshold requirements, the formatted data group is cross-compared with the dynamic information in the production line feedback mechanism to obtain an integrated unified data packet; the unified data packet is evaluated using an information distribution component to obtain comprehensive data.

[0047] Specifically, in the quality control scenario of a garment production line, the application of data extraction tools is a crucial step for real-time transmitted inspection information. The main function of data extraction tools is to separate key quality-related fields from complex information flows. For example, for the inspection information of a black suit, the information flow may contain multiple fields such as fabric texture, seam density, and wrinkle distribution. The data extraction tool will prioritize filtering out seam density and wrinkle distribution data directly related to quality, forming a quality field set. Assuming the total data volume of the information flow is 10MB, the extracted quality field set may only be 3MB, containing core data units such as seam anomaly points and concentrated wrinkle areas. For the report data generation target, template matching tools are used to map the core data units to a preset report format. Taking the inspection of the cuff of a black suit as an example, the core data unit includes data on cuff seam anomaly points. The template matching tool will map this data to the "Cuff Quality Assessment" section in the report format, forming a formatted data group. Assuming the report format requires each column to have no more than 500KB of data, the tool will automatically adjust the data layout to ensure compliance with structural requirements. Under the goal of information integration, the role of data fusion tools is to cross-reference formatted data sets with dynamic information from the production line feedback mechanism. Taking the suit collar area as an example, the formatted data set records the distribution of collar wrinkles, while the production line feedback mechanism may provide real-time parameters of the collar ironing equipment. The data fusion tool combines the two to generate a unified data package containing quality data and equipment status. Assuming that collar wrinkle data accounts for 60% and equipment parameters account for 40%, the fused data package can comprehensively reflect the correlation between quality and production status.

[0048] Furthermore, obtaining garment wrinkle detection results includes:

[0049] Based on the comprehensive data, key fields related to wrinkles are separated, a set of key fields is obtained, the set of key fields is matched with a preset tolerance standard, and dynamic correction is made based on the matching results to obtain the adjusted combination of acquisition parameters.

[0050] The adjusted combination of collected parameters is cross-compared with the production parameters on the production line to obtain an integrated parameter dataset; the integrated parameter dataset is then grouped to obtain garment wrinkle detection results.

[0051] Specifically, in the quality control scenario of garment production lines, the application of data extraction tools is crucial for comparing the range of wrinkles. The principle of data extraction tools lies in filtering fields related to specific quality indicators from the real-time transmitted information stream. Taking the inspection of a white shirt as an example, the information stream may contain information such as fabric color, texture distribution, and wrinkle location. The tool will prioritize extracting fields related to wrinkles, such as the percentage of wrinkle area and distribution density, forming a set of key fields. Assuming the total data volume of the information stream is 8MB, the extracted set of key fields is only 2.5MB, containing the core wrinkle data. The comparison of the key field set with preset tolerance standards can be implemented from the perspective of data matching. The comparison tool will compare the extracted wrinkle data with the standard value item by item. For example, if the standard stipulates that the wrinkle area percentage should not exceed 5%, while the actual data for the chest area of ​​the shirt is 6%, it is judged to exceed the tolerance range. This comparison process intuitively reflects the quality deviation and provides a basis for subsequent adjustments. Under the goal of adjusting the parameters of image acquisition equipment, the application of parameter configuration tools is reflected in the dynamic correction of acquisition settings. Taking shirt cuff detection as an example, if the wrinkle data exceeds the standard, it may be due to poor lighting angle during image acquisition, leading to recognition errors. The parameter configuration tool will adjust the device's light intensity and lens focal length to form a new combination of acquisition parameters, such as increasing the light intensity from 500 lumens to 600 lumens, ensuring improved image clarity and providing a more accurate data basis for subsequent detection.

[0052] This invention discloses an image analysis-based method for detecting garment wrinkles. It involves pre-establishing a multi-scene image database, acquiring and labeling garment image data under different materials and lighting conditions to obtain an initial training dataset. A convolutional neural network model is used to extract features and perform multi-scale analysis on the image data, resulting in a feature map containing subtle features. Noise interference caused by lighting variations is filtered out, and an image segmentation algorithm is used to initially locate the wrinkle region. An edge enhancement algorithm is then used to optimize the region boundaries. Finally, a matching analysis is performed using a pre-defined wrinkle shape template to obtain the final wrinkle detection result. This invention achieves accurate detection of garment wrinkles in complex environments and can be synchronized with the production line in real time, automatically generating detection reports, providing an effective solution for garment production quality control.

[0053] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for detecting garment wrinkles based on image analysis, characterized in that, include: Acquire clothing image data, perform annotation processing on the clothing image data, and obtain an initial training dataset; Feature extraction is performed on the initial training dataset to obtain a feature map; The feature map is filtered to obtain a denoised feature map; The denoised feature map is divided into regions to obtain a preliminary segmented folded region; The boundaries of the initially segmented folded regions are optimized to obtain clear fold boundary data; the clear fold boundary data is then combined with a preset fold shape template for matching analysis to obtain the final fold detection result. The final wrinkle detection result is synchronized and matched with the image acquisition frequency of the production line to obtain real-time transmitted detection information; Based on the real-time transmitted testing information, testing report data is automatically generated, and information is integrated with the production line feedback mechanism to obtain comprehensive data; The comprehensive data is correlated with production parameters to obtain garment wrinkle detection results; Obtaining the feature map includes: The clothing image data in the initial training dataset is decomposed into layers, and the decomposed images are divided into regions to obtain a set of layered image data. Using the layered image data set, multi-scale features are extracted using an edge detection tool. If the detected feature intensity is lower than a preset threshold, the clothing image data is enhanced to obtain an intermediate feature set containing subtle features. Based on the intermediate feature set, the features are integrated, and features at different scales are weighted and combined to obtain a unified feature representation map; the unified feature representation map is used to distinguish image content, and combined with the data feature analysis results, a feature map is obtained; Obtaining clear fold boundary data includes: Based on the denoised feature map, the pixel gradients around the region are analyzed to extract the key parts of the blurred boundary and obtain preliminary boundary information. Using the preliminary boundary information, the pixel values ​​of the boundary part are processed to enhance the contrast of the boundary region and obtain enhanced boundary feature data. Based on the enhanced boundary feature data, noise within the boundary area is cleaned up to obtain cleaned boundary lines; the continuity of the boundary lines is detected using the cleaned boundary lines, and interrupted or blurred parts are repaired to obtain clear folded boundary data.

2. The image analysis-based garment wrinkle detection method as described in claim 1, characterized in that, Obtaining the initial training dataset includes: Acquire clothing image data containing multiple scene images, classify and organize the clothing image data, perform preliminary noise reduction processing on the clothing image data, and obtain a clear first image set; Each image in the first image set is processed to distinguish regions, identify wrinkled and non-wrinkled regions in the images, and mark images whose detected edge features meet a preset threshold as wrinkled regions to obtain the initial training dataset.

3. The image analysis-based garment wrinkle detection method as described in claim 1, characterized in that, The denoised feature maps obtained include: Based on the feature map, the image content is divided into regions, the parts affected by noise are separated, and a preliminary set of feature regions is obtained. Based on the initially defined set of feature regions, brightness standardization is performed on each region. If the detected grayscale deviation exceeds a preset threshold, brightness correction is performed on the region to determine the corrected set of feature regions. Based on the corrected feature region set, noise interference is suppressed while preserving the integrity of subtle features, resulting in a denoised feature map.

4. The image analysis-based garment wrinkle detection method as described in claim 1, characterized in that, The final wrinkle detection results include: Based on the clear fold boundary data, preliminary extraction of morphological features is performed. Combined with a preset morphological template library, the morphological matching degree of the boundary data is compared to obtain a preliminary comparison result set. The regional lines in the preliminary comparison result set are supplemented to obtain an adjusted boundary dataset. Based on the adjusted boundary dataset, the detection accuracy value is checked a second time. If the detection accuracy value is lower than the preset threshold, the boundary data is optimized to determine the optimized boundary feature set. The boundary clarity is finally verified using the optimized boundary feature set. The correction data is then integrated using an image fusion tool to obtain the final wrinkle detection result.

5. The image analysis-based garment wrinkle detection method as described in claim 1, characterized in that, The real-time transmitted detection information includes: The acquired image data and the final wrinkle detection result are synchronously mapped on the time axis to obtain a time-consistent data set after matching; the redundant information in the detection result is simplified through the time-consistent data set, and the simplified data is hierarchically packaged to obtain a simplified data package; The simplified data packet is sent in segments, and the data integrity of the sent simplified data packet is checked to determine the complete data unit in the transmission process. If the complete data unit passes the check, the segmented data is recombined, and the recombined detection information is pushed by the information distribution component to obtain the real-time transmission detection information.

6. The image analysis-based garment wrinkle detection method as described in claim 1, characterized in that, The comprehensive data obtained includes: Based on the real-time transmitted detection information, key quality-related fields are separated from the information stream, a set of separated quality fields is obtained, and the core data units in the set are determined. The core data units are mapped to a pre-established report format to obtain the mapped formatted data group, and the data content that conforms to the report structure is determined. If the formatted data group meets the preset threshold requirements, the formatted data group is cross-compared with the dynamic information in the production line feedback mechanism to obtain an integrated unified data packet; the unified data packet is evaluated using an information distribution component to obtain comprehensive data.

7. The image analysis-based garment wrinkle detection method as described in claim 1, characterized in that, The results of garment wrinkle detection include: Based on the comprehensive data, key fields related to wrinkles are separated, a set of key fields is obtained, the set of key fields is matched with a preset tolerance standard, and dynamic correction is made based on the matching results to obtain the adjusted combination of acquisition parameters. The adjusted combination of collected parameters is cross-compared with the production parameters on the production line to obtain an integrated parameter dataset; the integrated parameter dataset is then grouped to obtain garment wrinkle detection results.

Images