A method and equipment for analyzing garment manufacturing processes

Abstract

This application discloses a method and apparatus for analyzing garment manufacturing processes. The method includes: extracting multi-level manufacturing process features of the garment structure from a garment pattern diagram uploaded by the client; constructing a garment structure semantic graph reflecting the garment structure hierarchy and manufacturing relationships based on the multi-level manufacturing process features; when the garment order is in a multi-size customization mode, if the garment order does not have pattern diagrams for other sizes, extracting grading process parameter information for other sizes from the garment order; performing grading deduction on the relevant node parameters in the garment structure semantic graph based on the grading process parameter information to obtain garment structure semantic graphs for other sizes; and extracting hierarchical features of the garment from the garment structure semantic graph and the garment structure semantic graphs for other sizes respectively to generate a multi-size manufacturing process parameter description of the garment. This improves the efficiency of garment manufacturing process analysis.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method and equipment for analyzing garment manufacturing processes. Background Technology

[0002] As consumer demand for personalization and fit continues to rise, clothing customization, especially multi-size bulk customization, is evolving from high-end, small-scale to a broader and more efficient production model.

[0003] When faced with a custom order, it's typically necessary to create a separate pattern drawing for each size. However, creating a pattern drawing for each size results in extended production time for the client. Furthermore, when analyzing multi-size garment customization, manufacturers must independently analyze each pattern drawing file to extract process parameters, consuming significant resources. Additionally, the timeline from order receipt to completion of process analysis and issuance of production instructions is significantly lengthened, especially when custom orders are fragmented (i.e., small order quantities per size but a wide size range), leading to low overall delivery efficiency. Summary of the Invention

[0004] To address the aforementioned problems, this application proposes a method for analyzing garment manufacturing processes, including: From the garment pattern drawings uploaded by the client, multi-level process features reflecting the garment structure are extracted; the multi-level process features include macroscopic structural features reflecting the garment's outline structure, component geometric features, and microscopic process features reflecting process details. Based on the aforementioned multi-level process features, a garment structure semantic graph reflecting the garment structure hierarchy and process relationships is constructed. When the clothing order is customized in a multi-size customization mode, if the clothing order does not have clothing pattern drawings for other sizes, extract the grading process parameter information for other sizes from the clothing order. Based on the grading process parameter information, the relevant node parameters in the garment structure semantic graph are graded and deduced to obtain garment structure semantic graphs for other sizes. The hierarchical features of the garment are extracted from the garment structure semantic map and the garment structure semantic maps of other sizes, respectively, to generate a multi-size process parameter description of the garment.

[0005] In one example, the construction of a garment structure semantic graph reflecting the garment structure hierarchy and process relationships based on the multi-level process features specifically includes: Corresponding graph nodes are established for the macroscopic structural features, component geometric features, and microscopic process features, respectively; the macroscopic structural features correspond to the root node, the component geometric features correspond to the child nodes, and the microscopic process features correspond to the leaf nodes; Based on the technological relationships between the features in the garment pattern drawing, edges are established between the nodes in the drawing; the edges include the membership edges connecting the root node and the child node, the assembly edges connecting the child nodes, the assembly edge relationships including at least one of the sewing relationship, nesting relationship and dependency relationship, and the attachment edges connecting the child node and the leaf node. The parameter information in the multi-level process features is assigned to the corresponding graph nodes as node attributes, thus obtaining the semantic graph of the garment structure.

[0006] In one example, based on the grading process parameter information, grading deduction is performed on the relevant node parameters in the garment structure semantic graph to obtain garment structure semantic graphs for other sizes, specifically including: Identify the target node corresponding to the grading process parameter information in the garment structure semantic graph; If the target node is a macroscopic structure node, query the associated component nodes connected to the target node from the clothing structure semantic graph; Based on the geometric relationship between the target node and the associated component node in the garment structure semantic graph, the grading process parameter information is converted into an equivalent displacement of the associated component node. If the target node is a component node or a micro-process node, the grading process parameter information is converted into a direct displacement amount of the component node or the micro-process node. Based on the hierarchical structure of the garment structure semantic graph, the equivalent displacement and / or the direct displacement, garment structure semantic graphs for other sizes are reconstructed.

[0007] In one example, the step of converting the grading process parameter information into an equivalent displacement of the associated component node based on the geometric relationship between the target node and the associated component node in the garment structure semantic graph specifically includes: Based on the geometric relationship between the target node and the associated component nodes in the garment structure semantic graph, the allocation rule between the size change of the target node and the displacement of each associated component node is determined; According to the allocation rule, the dimensional change in the grading process parameter information is allocated to the equivalent displacement of each associated component node.

[0008] In one example, the method further includes: Extract the first numerical sequence of the same target node under different sizes from the garment structure semantic graph and the garment structure semantic graphs of other sizes; Identify spatially adjacent associated nodes to the target node and extract the second numerical sequence of associated nodes under different sizes; For the first numerical sequence and the second numerical sequence, respectively, a function is fitted with size as the independent variable and parameter value as the dependent variable to obtain a first fitting function and a second fitting function. Analyze the mathematical relationship between the first fitting function and the second fitting function, and calculate the similarity that characterizes the consistency of the relationship between the first fitting function and the second fitting function; The similarity is compared with a preset similarity threshold to determine the parameter coordination between the target node and the associated node as the size changes, thereby completing the verification of the cross-size process coordination of the garment.

[0009] In one example, after comparing the similarity with a preset similarity threshold to determine the parameter consistency between the target node and associated nodes as size changes, the method further includes: Identify component pairs with stitching, nesting, and dependency relationships from the garment structure semantic map; From the garment structure semantic map and the garment structure semantic maps of other sizes, extract the relative position parameters of the component pairs under different sizes; the relative position parameters include at least one of spatial distance parameters, angle deviation parameters, and overlapping area parameters; The relative position parameters of the component pairs in the garment structure semantic diagram are used as constraints. Based on the relationship type and constraints of the component pairs, the relative position parameters under different sizes are analyzed to identify the assembly coordination of the component pairs as the size changes.

[0010] In one example, when the garment pattern drawing is in bitmap format, multi-layered technological features reflecting the garment structure are extracted from the garment pattern drawing uploaded by the client. These features include: The garment pattern diagram is decomposed into multiple scales to obtain images at different scale levels. The scale levels include macro scale, meso scale, and micro scale. The resolution of the macro scale is lower than that of the meso scale, the resolution of the meso scale is lower than that of the micro scale, and the resolution of the micro scale is the same as that of the garment pattern diagram. The overall outline features and structural segmentation lines of the garment are extracted from the macro-scale hierarchical image to obtain macro-structural features. The geometric features and relative positional relationships of garment components are extracted from the meso-scale hierarchical image to obtain component geometric features. The process details are extracted from the micro-scale hierarchical image to obtain micro-process features. The process details include at least one of sewing symbols, dart markings, and knife edge markings.

[0011] In one example, hierarchical features of the garment are extracted from the garment structure semantic map and the garment structure semantic maps of other sizes, respectively, to generate a multi-size process parameter description of the garment, specifically including: According to the hierarchical structure of the garment structure semantic graph, the garment structure semantic graph and the garment structure semantic graphs of other sizes are traversed layer by layer in the order from the root node to the leaf node. In each level of traversal, extract the node identifier, node attributes, and edge relationship information between the current level node and its child nodes; According to the hierarchical structure, the extracted node identifiers, node attributes, and edge relationship information are used as process parameter descriptions for each size.

[0012] In one example, the method further includes: When the clothing order is customized in a single-size customization mode, the hierarchical features of the clothing are extracted from the clothing structure semantic graph to generate a description of the clothing's process parameters.

[0013] On the other hand, embodiments of this application provide a garment process analysis device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform a garment process analysis method as described above.

[0014] The above-described technical solutions adopted in the embodiments of this application can achieve the following beneficial effects: By abstracting garment pattern diagrams into garment structure semantic graphs with a clear hierarchical structure, a formal expression of garment structure and process information is realized, enabling computers to understand garments and laying the data structure foundation for subsequent automated processing.

[0015] By using grading deduction based on the semantic graph of garment structure, the efficient generation of process parameters from single-size patterns to multi-size patterns is achieved. This avoids the repetitive work of drawing and analyzing pattern diagrams separately for each size, greatly saving time and computing resources, and thus improving the efficiency of garment customization production.

[0016] In summary, this significantly improves the level of intelligence in multi-size clothing customization scenarios. Attached Figure Description

[0017] To more clearly illustrate the technical solution of this application, some embodiments of this application will be described in detail below with reference to the accompanying drawings, in which: Figure 1 A flowchart illustrating a garment process analysis method provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of a garment process analysis device provided in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] Some embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0020] Figure 1 This is a flowchart illustrating a garment process analysis method provided in an embodiment of this application. The process can be executed by computing devices in a relevant field (such as a garment production service system), and certain input parameters or intermediate results in the process can be manually adjusted to help improve accuracy.

[0021] The analysis method involved in the embodiments of this application can be implemented by a terminal device or a server, and this application does not impose any special limitations on it. For ease of understanding and description, the following embodiments are all described in detail using a server as an example.

[0022] It should be noted that the server can be a single device or a system composed of multiple devices, i.e., a distributed server. This application does not make any specific limitations on this.

[0023] Figure 1 The process includes the following steps: S101: Extract multi-level process features reflecting the garment structure from the garment pattern drawing uploaded by the client; the multi-level process features include macroscopic structural features reflecting the garment outline, component geometric features, and microscopic process features reflecting process details.

[0024] A garment pattern drawing is a design document that illustrates the two-dimensional structure of a garment. It typically includes outlines, structural lines, component graphics, and technical details. Users submit garment pattern drawings from the client to the server, making it suitable for remote, collaborative modern garment production models.

[0025] Macro-structural features include the overall external outline of the garment, outline size information, structural dividing lines, and the geometric position information of structural dividing lines. For example, the external outline includes garment length, bust, waist, hip, and shoulder width, while structural dividing lines include princess seams, back seams, and side seams.

[0026] Component geometric features include the geometric attributes of each individual component and their relative positional relationships. For example, components include the front panel, back panel, sleeve panel, collar panel, pocket, waistband, shoulder gaiter, and facing. Geometric attributes include the outer contour shape, key point coordinates, side length, angle, curvature, and area of ​​each component. Positional relationships include the relative distance, alignment, angular relationships, and overlapping areas between components.

[0027] Microscopic process features include detailed information such as sewing symbols, dart markings, and cutting edge markings. For example, the type of sewing stitch (plain stitch, overlock stitch, coverstitch), alignment cross markings, the position and direction of pleats, the position and opening of darts (bust dart, waist dart, shoulder dart), and the position and direction of cutting edges.

[0028] When the garment pattern image is in bitmap format, the image undergoes preprocessing, including grayscale conversion, binarization, and denoising. Subsequently, the image is decomposed into multiple scales. For example, the original garment pattern image is defined as the microscale. It is downsampled to obtain a lower-resolution mesoscale image, and then downsampled again to obtain an even lower-resolution macroscale image. For instance, the mesoscale image has half the resolution of the original garment pattern image, and the macroscale image has one-quarter the resolution.

[0029] On macroscopic images, the overall outline and main structural dividing lines of the garment can be extracted using methods such as contour detection. On mesoscopic images, the geometric shape and relative position of each individual component can be extracted using methods such as connected component analysis, edge detection, and corner detection. On microscopic images, process details such as sewing symbols, dart markings, and cutting edge markings can be extracted using methods such as template matching, morphological processing, and character recognition.

[0030] S102: Based on the multi-level process features, construct a garment structure semantic graph that reflects the garment structure hierarchy and process relationships.

[0031] A garment structure semantic graph is a directed acyclic graph (DAG) or tree-structured data model. Its core lies in explicitly expressing the hierarchical and technological relationships between garment components through the definition of nodes and edges.

[0032] In the semantic graph of the garment structure, macroscopic structural features are root nodes, component geometric features are child nodes, and microscopic process features are leaf nodes.

[0033] Based on this, a semantic graph of garment structure reflecting the hierarchical structure and technological relationships of garments is constructed, including the following steps: Step 1: Establish corresponding graph nodes for the macroscopic structural features, component geometric features, and microscopic process features, respectively.

[0034] The root node corresponds to a macroscopic structural feature. Each macroscopic structural feature is mapped to a root node. The root node attributes store the identifier, size, and position parameters of the structural feature.

[0035] Child nodes correspond to component geometric features, with each component geometric feature mapped to a child node. Child nodes are connected to their root node via directed edges, indicating that component nodes influence macroscopic feature nodes. Child node attributes store the component's identifier, geometric parameters, material properties, etc.

[0036] Leaf nodes correspond to micro-level process features, with each micro-level process feature mapped to a leaf node. Leaf nodes are connected to their respective component nodes via directed edges, indicating that process details are attached to that component. Leaf node attributes store process type, specific parameters, operation instructions, etc.

[0037] Step 2: Based on the technological relationships between the features in the garment pattern drawing, establish edges between the nodes in the drawing; the edges include the membership edges connecting the root node and the child nodes, the assembly edges connecting the child nodes, the assembly edge relationships including at least one of the sewing relationship, nesting relationship and dependency relationship, and the attachment edges connecting the child nodes and the leaf nodes.

[0038] It's important to note that a stitching relationship refers to establishing a seam edge between two component nodes if they are marked as being stitched together in the pattern drawing. The edge attribute records the seam type and seam allowance. A nested relationship involves establishing a nested edge if one component node (e.g., a patch pocket) is located inside another component node (e.g., the garment body) and there is a boundary dependency. The edge attribute records the relative positioning method. A dependency relationship refers to establishing a dependency edge if the position, size, or existence of one component functionally depends on certain characteristic parameters of another component. For example, the length and width of a pocket flap must depend on the opening size of the pocket fabric. If the pocket fabric becomes larger, the pocket flap must be adjusted accordingly to avoid overclosing. Similarly, the number and position of belt loops depend on the width and wearing method of the belt. If the belt becomes wider, the width of the belt loops must increase, and the spacing between the loops may need to be adjusted.

[0039] Step 3: Assign the parameter information in the multi-level process features to the corresponding graph nodes as node attributes to obtain the garment structure semantic graph.

[0040] It should be noted that the semantic graph of clothing structure is stored in the form of a graph database or an in-memory adjacency list, which supports subsequent operations such as node query, parameter modification, and subgraph copying.

[0041] S103: When the customization mode of a garment order is multi-size customization mode, if the garment order does not have garment pattern drawings for other sizes, extract the grading process parameter information for each of the other sizes from the garment order.

[0042] Order information usually follows the pattern. Figure 1 The order can be uploaded or specified by the user through the client interface. For example, if the size field in the order contains multiple size values ​​(such as S, M, L, XL), and the field for whether to provide a full range of sizes is not specified, then the order will proceed to multi-size customization mode.

[0043] Grading process parameters are a set of rules used to describe the scaling or geometric transformation of sizes from garment pattern drawings to other sizes. Grading process parameters include a grade difference table, grading point displacement vectors, scaling factors, and rule expressions.

[0044] S104: Based on the grading process parameter information, perform grading deduction on the relevant node parameters in the garment structure semantic graph to obtain garment structure semantic graphs for other sizes.

[0045] The core of coding derivation lies in the fact that it does not perform affine transformations or free deformations on the garment pattern diagram, but rather performs mathematical operations on the parameter values ​​stored in the nodes of the garment structure semantic graph, and reasonably transmits the operation results to the relevant nodes based on the hierarchical relationships and geometric constraints in the garment structure semantic graph.

[0046] Specifically, determine the target object affected by the grading process parameters. For example, if the grading process parameter is a chest circumference increment of 4cm, then the target object is the macroscopic structural node of the chest circumference attribute. If the grading process parameter is shoulder end point displacement, then the target object is the coordinates of the key point representing the shoulder end point in the component node.

[0047] For parameters that directly affect node properties, update the node's property value directly. For example, add the displacement vector to the shoulder endpoint coordinates of a component node.

[0048] For parameters acting on macroscopic structural nodes, since these nodes do not directly correspond to any specific geometric entity, the changes in macroscopic dimensions must be allocated or converted into equivalent displacements of lower-level component nodes based on the connection relationships in the garment structure semantic graph. For example, an increase in bust size will cause displacement in the width direction of the front, back, and side panels; an increase in garment length will cause longitudinal displacement of the hem and / or related components. Allocation rules can be derived from garment pattern diagram structural knowledge; for example, the increase in bust size can be proportionally allocated to the center front, center back, and side seams.

[0049] Based on the grading process parameters corresponding to a specific size, after updating the parameters of all relevant nodes in the garment structure semantic graph, a new garment structure semantic graph corresponding to that size is generated. This garment structure semantic graph has the exact same topological structure as the uploaded one (node ​​types and connections are completely identical), differing only in the specific numerical parameters of the nodes. For multiple other sizes, the above process is repeated to obtain garment structure semantic graphs for multiple different sizes.

[0050] S105: Extract the hierarchical features of the garment from the garment structure semantic map and the garment structure semantic maps of other sizes respectively, and generate a multi-size process parameter description of the garment.

[0051] Specifically, the semantic graph of the garment structure for each size is traversed, macroscopic structural features are extracted from the root node, component geometric features are extracted from the child nodes, and microscopic process features are extracted from the leaf nodes.

[0052] Based on this, the hierarchical features of the clothing are extracted, including the following steps: Step 1: Following the hierarchical structure of the garment structure semantic graph, traverse the garment structure semantic graph and the garment structure semantic graphs of other sizes layer by layer from the root node to the leaf node.

[0053] Step 2: In each level of traversal, extract the node identifier, node attributes, and edge relationship information between the current level node and its child nodes; Step 3: According to the hierarchical structure, the extracted node identifiers, node attributes, and edge relationship information are used as process parameter descriptions for each size.

[0054] It should be noted that the description of the multi-size process parameters of clothing is a structured dataset that integrates size dimensions. The typical form is: a multi-dimensional array or table organized by size index (e.g., S, M, L, XL), or a process sheet file generated independently for each size, with an attached size identifier.

[0055] The process parameter descriptions can be pushed to the production execution system or fed back to the user for confirmation through the client, thereby completing the fully automated parsing of the process from single-size pattern drawing to multi-size manufacturable process data.

[0056] It should be noted that when the clothing order is customized in a single-size customization mode, the hierarchical features of the clothing are extracted from the clothing structure semantic graph to generate a description of the clothing's process parameters.

[0057] It should be noted that, although the embodiments in this application are based on... Figure 1Steps S101 to S105 will be described sequentially, but this does not mean that steps S101 to S105 must be performed in a strict order. The reason this embodiment follows this order is... Figure 1 The order in which steps S101 to S105 are described is provided to facilitate understanding of the technical solutions of the embodiments of this application by those skilled in the art. In other words, in the embodiments of this application, the order of steps S101 to S105 can be appropriately adjusted according to actual needs.

[0058] pass Figure 1 The method abstracts garment pattern diagrams into garment structure semantic graphs with a clear hierarchical structure, realizing the formal expression of garment structure and process information. This enables computers to understand garments and lays the data structure foundation for subsequent automated processing.

[0059] By using grading deduction based on the semantic graph of garment structure, the efficient generation of process parameters from single-size patterns to multi-size patterns is achieved. This avoids the repetitive work of drawing and analyzing pattern diagrams separately for each size, greatly saving time and computing resources, and thus improving the efficiency of garment customization production.

[0060] In summary, this significantly improves the level of intelligence in multi-size clothing customization scenarios.

[0061] based on Figure 1 In addition to the method described herein, this application also provides some specific implementation schemes and extension schemes of the method, which will be further explained below.

[0062] In one example, a garment pattern drawing is a complex of multi-granular information. A single drawing includes centimeter-level overall outlines (such as garment length and bust), millimeter-level detailed manufacturing symbols (such as alignment seams and dart points), and geometric shapes of components in between (such as sleeves, collars, and pockets). Traditional image processing methods often process garment pattern drawings at a single resolution, resulting in low accuracy in garment feature extraction. For example, if the resolution is reduced to capture the macroscopic outline, microscopic details are lost. If the original high resolution is used to preserve microscopic details, macroscopic structural feature extraction is affected by noise and involves enormous computational costs.

[0063] Based on this, the multi-layered technological features reflecting the garment structure are extracted from the garment pattern drawings uploaded by the client, including the following steps: Step 1: Decompose the garment pattern drawing into multiple scales to obtain images at different scale levels. The scale levels include macro scale, meso scale, and micro scale; the resolution of the macro scale is lower than that of the meso scale, the resolution of the meso scale is lower than that of the micro scale, and the resolution of the micro scale is the same as that of the garment pattern drawing.

[0064] When the garment pattern drawing is in bitmap format (such as PNG, JPEG, TIFF), the original resolution of the uploaded file is used directly as the microscale image. The microscale image is downsampled to obtain the mesoscale image, for example, reducing the resolution to half of the original resolution. The mesoscale image is downsampled again to obtain the macroscale image, for example, reducing the resolution to half of the mesoscale resolution.

[0065] It should be noted that downsampling algorithms can include nearest neighbor interpolation, bilinear interpolation, bicubic interpolation, and region pixel averaging. To preserve the sharpness of the contour edges, region pixel averaging downsampling should be preferred.

[0066] Step 2: Extract the overall outline features and structural segmentation lines of the garment from the macro-scale hierarchical image to obtain macro-structural features; extract the geometric features and relative positional relationships of garment components from the meso-scale hierarchical image to obtain component geometric features; and extract the process details from the micro-scale hierarchical image to obtain micro-process features; process details include at least one of sewing symbols, dart markings, and knife edge markings.

[0067] Specifically, the macroscopic level image has the lowest resolution, preserving the large-scale structural information of the garment pattern while significantly suppressing high-frequency noise such as internal details and process symbols of the components.

[0068] The contour feature extraction process includes: First, converting the grayscale image into a black-and-white binary image to separate the clothing area (foreground) from the background. Global thresholding, adaptive thresholding, or edge detection-based threshold selection can be used. Then, the Suzuki algorithm is applied to extract the outer contour of the foreground region. This outer contour is the overall outer envelope of the clothing.

[0069] It's important to note that the Suzuki algorithm connects the pixels in a binary image sequentially to form a contour. For example, the process might include: First, scanning each pixel from top to bottom and left to right. Then, when encountering a foreground (clothing) pixel that has never been visited before, and its left side is the background, that point is marked as the starting point of the contour. Boundary tracing: Starting from the starting point, searching for the eight adjacent pixels in a clockwise (or counter-clockwise) direction. If the adjacent pixel is foreground, moving to it and recording its coordinates. If it's background, turning away. This rule ensures that the algorithm follows the boundaries. Finally, when returning to the starting point, a closed contour is formed.

[0070] Furthermore, to improve the efficiency of subsequent geometric calculations, the original contour point set can be simplified. The goal of simplification is to preserve the main shape and transition features of the contour while eliminating redundant intermediate points.

[0071] For example, the Douglas-Puk algorithm can be used to simplify the contour. By setting a distance threshold, points that deviate slightly from the main contour direction can be recursively deleted, while key vertices are retained.

[0072] It's important to note that the Douglas-Puk algorithm removes redundant points, retaining only key inflection points. For example, the process might include: Connecting the beginning and end: Draw a virtual straight line between the start and end points of the contour. Finding the farthest point: Calculate the distance from all points on the contour to this line. Find the point with the farthest distance. If this maximum distance is greater than a set threshold (e.g., 0.5 mm), it indicates a large bend that cannot be simplified, and this point is retained. Recursively segmenting: Divide the contour into two segments using this farthest point as the boundary. For each segment, repeat the steps of connecting the beginning and end and finding the farthest point. Termination: When the distance from all points to their corresponding lines is less than the threshold, the remaining points are the simplified contour vertices.

[0073] Since macroscopic structural features (such as garment length and chest circumference) typically depend on the extreme points of the contour, the coordinate distribution of the original contour point set can also be directly analyzed. For example, by traversing the row and column coordinates of the contour points, the coordinate values ​​of the leftmost, rightmost, topmost, and bottommost points can be directly obtained, thereby calculating parameters such as garment length and chest circumference. This method does not require contour simplification and directly uses the original contour data for extreme value calculation.

[0074] For example, the minimum bounding rectangle of the silhouette provides the garment length (rectangle height) and bust / hip circumference (approximate rectangle width, which needs to be adjusted according to the garment structure); the silhouette perimeter provides reference values ​​for hem circumference, neckline circumference, etc.

[0075] Specifically, structural dividing lines (such as princess lines, back seams, side seams, yoke lines, and shoulder lines) appear as continuous curves that run through the interior or boundaries of components on a macro-scale hierarchical image.

[0076] For example, prepare a large number of labeled clothing pattern sample images. During labeling, directly color different types of structural dividing lines with different colors (e.g., princess seams - red, side seams - blue, background - black). Model training: Train a semantic segmentation network (such as UNet) or an instance segmentation network (such as Mask R-CNN). Based on the clothing pattern sample images, teach the model which type of line each pixel belongs to.

[0077] Based on this, the garment pattern drawing is input into the trained model, which directly outputs a color image indicating which pixels represent princess seams and which represent side seams. Then, simple skeleton extraction or contour tracing is performed on the categorized pixels to obtain vectorized lines.

[0078] Specifically, the mesoscale layer images have moderate resolution, and individual components (front piece, back piece, sleeve piece, collar piece, pocket, etc.) in the images have formed clear connected regions, and the relative positional relationships between components are preserved. At the same time, the fine noise at the microscale has been downsampled and smoothed, and is less likely to interfere with component segmentation.

[0079] The component geometric feature extraction process includes: First, the binarized mesoscopic image is labeled with connected components, and each independent closed region is labeled as a candidate component. Then, based on the geometric features of the connected regions, such as area, aspect ratio, rectangularity, and circularity, combined with prior knowledge of clothing patterns (e.g., the front piece usually has the largest area and includes a neckline cutout, and the sleeve piece is usually approximately trapezoidal or fan-shaped), the candidate components are classified and assigned component type labels (front piece, back piece, sleeve piece, collar piece, pocket, facing, etc.).

[0080] Finally, for each component region, calculate the geometric features.

[0081] It should be noted that geometric features may include the following: Outer contour polygon: Extract the sequence of contour points of the component region, simplify it using the Douglas-Puk algorithm, and use it as the shape description of the component.

[0082] Key point coordinates: Identify feature points on the component outline. For example, key points for a sleeve include the sleeve cap point, left end point of the cuff, right end point of the cuff, front armhole alignment point, and back armhole alignment point; key points for a collar include the collar center point, collar point, and collar stand height point. Key points can be located through corner detection, curvature extreme point detection, or template matching.

[0083] Dimensional parameters: Calculate the minimum bounding rectangle of the part to obtain the part's width and height; calculate the equivalent ellipse of the part's outline to obtain the major and minor axes; measure the Euclidean distance between two specific points (such as the distance from the shoulder end point to the hem point).

[0084] Area and perimeter: Accurately calculates the pixel area and outline perimeter of the component region.

[0085] The relative positional relationship extraction process includes: For spatial distance, calculating the Euclidean distance between the center points of two components; or calculating the shortest distance, vertical distance, and horizontal distance between the boundaries of two components. For angular relationships, calculating the angle between the principal axes of two components; or calculating the angle between a certain edge of a component and the global horizontal or vertical direction. For alignment relationships, detecting whether two components are aligned in the horizontal or vertical direction. For example, the horizontal offset between the pocket center line and the front center line of the garment piece; whether the sleeve cap points of the left and right sleeve pieces are at the same height. For overlap relationships, calculating the intersection-union ratio of the two component regions.

[0086] Specifically, microscale hierarchical images retain the original highest resolution and are a reliable level for extracting fine process details.

[0087] The process of identifying and locating garment construction details can include: establishing a template library of construction symbols, covering various standard garment construction symbols such as darts, pleats, slits, buttonholes, and sewing stitch types. On a microscopic image, a sliding window approach is used to calculate the normalized cross-correlation (NCC) coefficient between each window and all templates in the template set. When the maximum correlation coefficient exceeds a preset threshold (e.g., 0.8), it is determined that a corresponding construction symbol exists at that location, and its coordinates and type are recorded.

[0088] For each identified process detail, specific parameters can be further quantified. For example: For darts: detect two converging straight lines and their intersection (dart tip), measure the dart opening width, dart length, and the angle between the dart and the component boundary. For alignment cuts: detect the position coordinates, length, and direction of short line segments (vertical cut, V-shaped cut). For buttonholes: detect the buttonhole outline, extract the center point coordinates, buttonhole length, diameter, number of buttonholes, and spacing. For pleats: detect pleat symbols (usually Z-shaped or wavy lines), measure pleat depth, pleat spacing, and pleat direction. For sewing stitches: detect the start, end, and turning points on the stitch trajectory, calculate the total stitch length, and identify the stitch type (single-needle plain stitch, double-needle coverstitch, three-thread overlock, etc.).

[0089] It should be noted that garment pattern drawings may include text labels, such as seam allowance 1cm, alignment here, dart allowance 2.5cm. Therefore, a region of interest (ROI) can be defined, and optical character recognition can be used to extract the text content and parse the numerical values ​​and units within it.

[0090] Furthermore, the identification of process details can also be achieved through end-to-end deep learning models (such as LayoutLM, multi-task CNN+RNN). This model adopts an encoder-decoder architecture, with microscale images as input and a series of structured process information entities as output. Each entity includes its bounding box, category label (such as provincial highway symbols, text annotations), and text content (if it is text).

[0091] In summary, at the macroscopic scale, downsampling suppresses isolated pixel noise generated by microscopic process symbols, ensuring that overall contour detection is unaffected by tool marks, sprue lines, etc. At the microscopic scale, the original resolution is preserved, ensuring that the accuracy of process detail localization is not affected by pixel loss. This strategy of equipping observation windows with different resolutions for information of different granularities fundamentally solves the dilemma of detail preservation and noise suppression under a single resolution. This results in a significant improvement in the signal-to-noise ratio of feature extraction.

[0092] Macro-scale images require less data than the original garment pattern image, resulting in a speedup of several to tens of times for contour detection algorithms at this scale. Meso-scale images also require less data than the original garment pattern image, significantly improving the efficiency of component segmentation and geometric calculations. Multi-scale decomposition reduces the overall computational load compared to performing the full algorithm directly on the original image. This achieves a significant optimization in computational efficiency.

[0093] In one example, when the garment pattern drawing is a vector file (such as DXF, AAMA, or TIIP files), the process of extracting multi-layered technological features reflecting the garment structure from the garment pattern drawing uploaded by the client includes the following steps: Step 1: Parse the file structure of the garment pattern drawing, obtain the primitive objects, and identify the closed paths used to describe the overall outline of the garment from the primitive objects as macroscopic structural features.

[0094] The file format is identified, and the corresponding parsing library (e.g., an open-source or commercial CAD parsing library) is called to read the file content. After parsing, a series of primitive objects are obtained, each of which includes geometric definitions (such as lines, arcs, splines, polylines, block references, etc.) and additional attributes (such as layers, colors, linetypes, extended data, etc.).

[0095] In vector files, the overall outline is typically composed of one or more closed outer polylines and is often located on a specific layer (e.g., the outline layer). The system iterates through all primitives and filters out closed paths that meet the following conditions: located on a specified layer, closed, and with the largest area (or located on the outermost layer). These are identified as the overall outer outline of the garment. Then, by calculating the minimum bounding rectangle, area, perimeter, and other parameters of this outline, macroscopic dimensions such as garment length, bust, and hem are directly obtained.

[0096] Step 2: Identify the block reference or closed polyline corresponding to the independent clothing component from the primitive object, and determine the component type and geometric parameters according to the preset block name, layer name or extended attributes, as the component geometric features.

[0097] The geometric features of a component correspond to the individual cut pieces that make up the garment (such as the front piece, back piece, sleeve pieces, pockets, etc.). In a vector file, each individual component is typically defined as an independent block reference or closed polyline, with a clear naming convention. By recognizing these block references or polylines and reading their names, layers, and geometric vertex coordinates, the system can obtain the precise outline of the component.

[0098] Furthermore, based on preset component type mapping rules, block names (e.g., FrontLeft, Sleeve) or layer names (e.g., garment layer, sleeve layer) are mapped to semantic component type labels (e.g., front garment piece, sleeve piece). Simultaneously, geometric calculations are used to obtain the component's width, height, key point coordinates (e.g., shoulder point, sleeve top point), area, and relative positional relationship with other components.

[0099] Step 3: Identify labeled entities, text entities, and specific process symbol entities from the graphic object, and read their positions, text content, and associated attributes as micro-process features.

[0100] In vector files, micro-scale process features are typically represented by specific primitives.

[0101] Process symbols, such as darts, pleats, and buttonholes, are often defined as block references or custom entities with specific attributes. The system identifies the type of process symbol by recognizing the name (e.g., Dart, Button) or extended attributes of these block references, and obtains the positioning coordinates, rotation angle, and dimensional parameters.

[0102] Text annotations, such as seam allowance, size markings, and alignment instructions, exist in the form of text entities. The system reads the text content, position coordinates, font style, and other information, and parses the values ​​and units using regular expressions or natural language processing technology (for example, parsing the parameter name "gutter allowance", the value "2.5", and the unit "cm" from "gutter allowance 2.5cm").

[0103] Association Relationships: In vector files, text annotations are often associated with corresponding process symbols or components (e.g., through group codes or extended data links). The system can utilize this association information to directly bind the text parsing results to the corresponding process symbol or component nodes, forming a complete description of the microscopic process features.

[0104] In one example, traditional systems can only understand the coordinate changes of a specific point, and cannot directly map macroscopic dimensional changes. If pattern makers assign displacement vectors to each grading point based on macroscopic dimensional changes, the work efficiency is low and errors are easy to occur.

[0105] Based on this, the relevant node parameters in the garment structure semantic graph are subjected to encoding and deduction, including the following steps: Step 1: Identify the target node corresponding to the grading process parameter information in the garment structure semantic graph.

[0106] There is a clear semantic mapping relationship between the grading process parameters and the nodes of the garment structure semantic graph, which can establish a mapping rule base for the garment structure semantic graph.

[0107] Based on this, the mapping rule library is queried according to the grading process parameters to determine the corresponding target node. For example, if the garment length difference is 2cm, the target node is the garment length.

[0108] Step 2: If the target node is a macroscopic structure node, query the associated component nodes connected to the target node from the clothing structure semantic graph.

[0109] The query identifies component nodes that are directly connected to the target macrostructure node. These component nodes are child nodes connected to the root node of the macrostructure via directed edges. When the attributes of the macrostructure node (root node) change, all its directly connected child nodes become associated component nodes.

[0110] It should be noted that graph traversal algorithms (such as breadth-first search (BFS) and depth-first search (DFS)) are used to start from the macroscopic structure nodes and traverse the directly connected component nodes downwards along the directed edges to form a set of associated component nodes.

[0111] Step 3: Based on the geometric relationship between the target node and the associated component node in the garment structure semantic graph, the grading process parameter information is converted into an equivalent displacement of the associated component node.

[0112] Geometric relationships refer to the quantitative dependencies between macroscopic structural node dimensional parameters and component node parameters. A solution method based on geometric constraints is employed. For example, the algebraic relationship between macroscopic structural node dimensional parameters and component node parameters is extracted from the semantic graph of the garment structure. The variation in macroscopic structural node dimensional parameters is set, and the displacement of key points of each associated component node is used as the unknown, thereby solving for the component node displacement.

[0113] It should be noted that each associated component node may include multiple affected boundary key points, and each key point obtains an independent displacement vector.

[0114] Step 4: If the target node is a component node or a micro-process node, convert the grading process parameter information into a direct displacement of the component node or the micro-process node.

[0115] When the identified target node is a component node or a micro-process node, the grading parameter information itself already includes specific displacement commands or direct dimensional modification commands. Therefore, there is no need for complex equivalent conversions; instead, a direct displacement conversion is performed.

[0116] When the grading process parameters include displacement amount and displacement direction, this displacement vector is directly assigned to the corresponding key point coordinate attribute in the target node.

[0117] It should be noted that if only the displacement amount is specified in the grading process parameters but not the displacement direction (such as moving the shoulder point outward by 2mm), the displacement vector will be automatically determined based on the normal direction of the key point in the garment structure semantic diagram or industry conventions (such as the shoulder point usually moving outward along the shoulder slope direction).

[0118] When the grading process parameters include component size scaling parameters (such as enlarging the collar circumference by 105%), this scaling factor is parsed and applied to the contour vertex coordinates of the component nodes. Specifically, with the geometric center of the component (or a specified reference point) as the origin, the coordinates of all contour vertices are multiplied by the scaling factor.

[0119] When the grading process parameters are directly applied to the micro-process nodes, the displacement vector is applied to the coordinate attributes stored in the leaf nodes.

[0120] Step 5: Based on the hierarchical structure of the garment structure semantic graph, the equivalent displacement and / or the direct displacement, reconstruct and generate garment structure semantic graphs for other sizes.

[0121] The principle of reconstruction is: the topological structure remains unchanged. The semantic graphs of clothing structure for other sizes have the exact same set of nodes, node types, connections between nodes, and edge types and attributes as the original clothing structure semantic graph. The only thing that changes is the numerical parameters (coordinates, length, angle, area, etc.) stored in the nodes.

[0122] When the boundary key point coordinates of a component node are displaced, the following derived attributes of that node need to be automatically recalculated. For example, the vertex sequence of the outline polygon, component width, height, area, component center point coordinates, and the relative distance and angle between the component and other components.

[0123] Furthermore, based on the edge relationships and geometric constraints of the garment structure semantic graph, the displacement of associated component nodes can be considered to further affect other component nodes or micro-process nodes connected to them.

[0124] Specifically, based on the topology and edge relationships of the garment structure semantic graph, the affected nodes are traversed using a graph traversal algorithm. During the traversal, the equivalent displacement is transferred to the affected nodes according to the geometric constraints between nodes extracted from the garment structure semantic graph. The affected nodes include component nodes or micro-process nodes.

[0125] When the target node is a component node, the displacement of the component node can be further affected by the edge relationships and geometric constraints of the garment structure semantic graph, based on the edge relationships and geometric constraints of the component node.

[0126] Specifically, based on the topological structure and edge relationships of the clothing structure semantic graph, the affected nodes are traversed using a graph traversal algorithm. During the traversal, the direct displacement is transferred to the affected nodes according to the geometric constraints between nodes extracted from the clothing structure semantic graph.

[0127] It should be noted that the service order information allows users to define code transfer rules. Code transfer rules specify the transfer path nodes and transfer methods for code parameters. These transfer methods include position transfer, size transfer, and angle transfer. Position transfer means that the downstream node moves with the upstream node while its own geometry remains unchanged; size transfer means that the downstream node's geometry adjusts proportionally with the upstream node. The spatial orientation of the downstream node adjusts with changes in the upstream node's angle parameters (e.g., rotation), but its geometry and relative position remain unchanged.

[0128] For example, the transfer path nodes of the front garment piece are the pocket and the pocket flap in sequence: the transfer method between the front garment piece and the pocket is positional transmission (the pocket moves with the front garment piece), and the transfer method between the pocket and the pocket flap is size transmission and positional transmission (the pocket flap must move with the pocket and be adjusted according to the change ratio of the pocket).

[0129] Based on this, the equivalent displacement or direct displacement can also be transferred to the affected component nodes or micro-process nodes by combining the code transfer rules.

[0130] It should be noted that displacement transfer can be automatically triggered by the system's default geometric constraints (e.g., nodes on both sides of the seam edge must be displaced synchronously to maintain a match), or it can be customized by the user through the grading transfer rules defined in the service order information. When the user defines grading transfer rules, the user's rules will be executed first; otherwise, the transfer will be based on the default geometric constraints.

[0131] In summary, the system can automate the execution of macro-level grading rules based on structural knowledge and geometric relationships in the garment structure semantic graph. It can also automatically calculate the appropriate displacement for each node, eliminating the need for pattern makers to manually specify displacement vectors for each grading point, thus significantly improving grading efficiency.

[0132] Since the coding deduction is performed within the semantic graph framework and based on the hierarchical relationship and geometric constraints between nodes, the logical consistency between macroscopic dimensional changes and component displacements, and between component displacements and process detail displacements, is always maintained, ensuring the structural consistency of the coding deduction.

[0133] In one example, converting the grading process parameter information into an equivalent displacement for the associated component node includes the following steps: Step 1: Based on the geometric relationship between the target node and the associated component nodes in the garment structure semantic graph, determine the allocation rule between the size change of the target node and the displacement of each associated component node.

[0134] Geometric relationships include dimensional decomposition relationships and boundary definition relationships.

[0135] Dimensional decomposition relationship: The dimensions of macroscopic structural nodes are composed of the corresponding dimensional constraints of multiple components. For example, the bust circumference is composed of the sum of the width of the front piece and the width of the back piece. When the bust circumference increases, the increment needs to be allocated according to the original width ratio of the components.

[0136] Boundary definition relationship: Structural dividing lines (such as princess seams and side seams) define the common boundaries of adjacent components. When the position of the dividing line shifts (such as the princess seam moving towards the side seam), the boundaries of the two components on both sides need to be simultaneously displaced in opposite directions to maintain the stitching relationship. At this time, the displacement amounts are equal and the directions are opposite.

[0137] Step 2: According to the allocation rules, allocate the dimensional change in the grading process parameter information to the equivalent displacement of each associated component node.

[0138] For example, if the chest circumference increases by 5cm, and the front and back pieces each occupy 1 / 2 of the chest circumference, then the width of the front piece increases by 2.5cm, and the width of the back piece increases by 2.5cm.

[0139] In summary, by using geometrically driven allocation rules, the grading parameters of macroscopic structural nodes can be automatically and reasonably decomposed into the displacements of component nodes, thus avoiding pattern distortion after grading.

[0140] In one example, semantic maps of garment structures for other sizes were generated through grading. However, when multiple grading rules provided by the client have potential geometric contradictions, if the system cannot identify and warn of these contradictions, it can only execute them blindly, potentially resulting in unreasonable or even unsewn patterns.

[0141] Based on this, cross-size process coordination verification of clothing can also be performed based on the clothing structure semantic diagram and the clothing structure semantic diagrams of other sizes, including the following steps: Step 1: Extract the first numerical sequence of the same target node under different sizes from the garment structure semantic graph and the garment structure semantic graphs of other sizes.

[0142] Because the semantic graph structure remains consistent across different sizes (only the node parameter values ​​change), nodes can be indexed using the same node ID or path. From the garment structure semantic graph for each size, the parameter values ​​of the target node are extracted, and these extracted parameter values ​​are organized into a sequence according to size order, serving as the first value sequence.

[0143] Step 2: Identify the associated nodes that are spatially adjacent to the target node, and extract the second numerical sequence of the associated nodes under different sizes.

[0144] Based on the geometric relationships in the garment structure semantic map or garment structure semantic maps for other sizes, identify nodes that are spatially adjacent to the target node. For each associated node, extract values ​​from the garment structure semantic map for each size to generate a second numerical sequence.

[0145] It should be noted that the target node and associated node corresponding to the coding process parameters are spatially adjacent, indicating that they are connected by an edge.

[0146] Step 3: For the first numerical sequence and the second numerical sequence, perform function fitting with size as the independent variable and parameter value as the dependent variable respectively to obtain the first fitting function and the second fitting function.

[0147] The independent variable is size, which can be converted into numerical values, for example, S=1, M=2, L=3, XL=4.

[0148] It should be noted that the fitting model is selected based on the characteristics of the data, usually a linear model (such as a linear function) or a polynomial model. In garment grading, most parameters change linearly, so linear fitting can be given priority.

[0149] Mathematical methods such as least squares are used to fit the data points, and the coefficients of the fitted function are obtained. And b. For linear fitting, the function form is: .

[0150] Step 4: Analyze the mathematical relationship between the first fitting function and the second fitting function, and calculate the similarity that characterizes the consistency of the relationship between the first fitting function and the second fitting function.

[0151] The consistency of the relationship can be judged by residual analysis based on linear regression or by calculation based on derivative sequences. The smaller the residual, the more consistent the relationship between the two functions. The closer the correlation coefficient or cosine similarity of the two derivative sequences is to 1, the more consistent the relationship between the two functions.

[0152] Step 5: Compare the similarity with a preset similarity threshold to determine the parameter coordination between the target node and the associated node as the size changes, thereby completing the verification of the cross-size process coordination of the garment.

[0153] For example, if the sum of squared residuals is used to judge the consistency of a relationship, the similarity threshold is set to 0.1 (a residual less than 0.1 indicates harmony). If the correlation coefficient is used to judge the consistency of a relationship, the similarity threshold is set to 0.95 (a correlation coefficient greater than 0.95 indicates harmony).

[0154] If the similarity meets the threshold condition, the clothing is deemed to be consistent across different sizes, and the verification is passed.

[0155] If the conditions are not met, a coordination problem is identified, triggering an alert or marking the target node and its associated nodes as pending review.

[0156] In summary, by quantitatively analyzing the patterns and interrelationships of parameters as they change with size, we can intelligently determine the rationality and consistency of grading parameters, ensuring that the derived multi-size patterns are not only dimensionally correct but also have coordinated processes and are sewn.

[0157] Further, calculating the similarity between the first fitting function and the second fitting function includes the following steps: Step 1: Calculate using linear regression to find a proportionality coefficient and an offset that minimize the difference between the first and second fitted functions.

[0158] The first fitting function describes the pattern of target node variation with size, and the second fitting function describes the pattern of related node variation with size.

[0159] Specifically, for each size independent variable, the first and second fitted function values ​​are calculated respectively, resulting in multiple pairs of function values. Then, linear regression analysis is performed on the multiple pairs of function values, and the proportionality coefficient and offset are calculated using the least squares method.

[0160] Assume the second fitting function With the first fitting function There is a linear relationship between them: .in, Let be the scaling factor and b be the offset. The goal is to find the optimal... And b, so that the model predictions are similar to the actual values. The difference in values ​​is minimal.

[0161] The solution is obtained using the least squares method, and the expression is as follows:

[0162]

[0163] in, for The mean, for The mean.

[0164] Step 2: Calculate the coefficient of determination for the linear regression model and use the coefficient of determination as the similarity.

[0165] For each data point, calculate the residual between the predicted value and the actual value, as shown in the following expression:

[0166] The residual represents the degree to which a point deviates from the fitted line. This represents the residual corresponding to the i-th data point. Let represent the function value of the first fitting function at the i-th data point (i-th size). This represents the function value of the second fitting function at the i-th data point (i-th size).

[0167] The expression for the sum of squared residuals is as follows:

[0168] The sum of squared residuals is the overall deviation measure. Residuals can be positive or negative (positive when the point is above the line, negative when it is below). Directly adding them together will cancel out the positive and negative values, failing to reflect the overall deviation. Squaring makes all values ​​non-negative and amplifies the impact of larger residuals, making it more sensitive to large deviations. Summing up the squared residuals of all points yields an overall deviation measure, the sum of squared residuals.

[0169] The expression for the coefficients is as follows:

[0170]

[0171] in, To determine the coefficients, the closer they are to 1, the more consistent the relationship between the two functions. The closer they are to 0, the more likely the two functions are unrelated or have a weak relationship. TSS reflects the overall fluctuation of the second fitted function value with size variation, representing the sum of squares of the deviations from all data points.

[0172] Furthermore, calculating the similarity between the first fitting function and the second fitting function may further include the following steps: Step 1: Calculate the first derivative sequences of the first and second fitting functions for all sizes to obtain the first derivative sequence and the second derivative sequence.

[0173] For each size point, calculate the first derivative of the first and second fitted functions at that point. If the functions are linear, the derivatives are constants. If the functions are nonlinear, numerical differentiation methods can be used.

[0174] Arrange the derivative values ​​at all size points in order to obtain the derivative sequence. The derivative sequence reflects the rate of change of the parameter with size. If the rate of change of two nodes remains constant or equal across all sizes, it indicates that they are coordinated during the grading process.

[0175] Step 2: Calculate the cosine similarity between the first derivative sequence and the second derivative sequence, and use the cosine similarity as a measure of relationship consistency.

[0176] If two vectors are in the same direction, the cosine value is 1. Treating the derivative sequence as a vector, we use the cosine of their included angle to measure the similarity of their changing trends. The closer the value is to 1, the higher the consistency.

[0177] The expression is as follows:

[0178] in, This represents the i-th element of the first derivative sequence (corresponding to the derivative value of the i-th size). This represents the i-th element of the second derivative sequence (corresponding to the derivative value of the i-th size).

[0179] In one example, the coordination of garment craftsmanship is reflected not only in the variation of parameter values, but also in the spatial relationship between components.

[0180] Based on this, after comparing the similarity with a preset similarity threshold to determine the parameter coordination between the target node and the associated node as size changes, the following steps are included: Step 1: Identify component pairs with stitching, nesting, and dependency relationships from the garment structure semantic map.

[0181] Specifically, the semantic graph of the clothing structure is traversed to find all edges that satisfy the above relationship type. For each edge that meets the conditions, the two component nodes connected to it are extracted to form a component pair.

[0182] It should be noted that the relationship type of component pairs (such as stitching, nesting, dependency, etc.) and relationship attributes (such as stitching position, nesting depth, etc.) can be recorded simultaneously.

[0183] Step 2: Extract the relative position parameters of the component pairs under different sizes from the garment structure semantic map and the garment structure semantic maps of other sizes; the relative position parameters include at least one of spatial distance parameters, angle deviation parameters, and overlapping area parameters.

[0184] Among them, the spatial distance parameter reflects the straight-line distance and outline distance between two parts, the angle deviation parameter reflects the relative angle between two parts (such as the angle between the sleeve and the body), and the overlapping area parameter reflects the area or length of overlap between two parts (such as the amount of overlap in the seam area).

[0185] Specifically, for each component pair, in the garment structure semantic graph for each size: locate the geometric nodes of the first and second components, extract the corresponding relative position parameters, and sort the extracted parameter values ​​by size.

[0186] Step 3: Use the relative position parameters of the component pairs in the garment structure semantic map as constraints.

[0187] From the semantic diagram of garment structure, the relative positional parameters of each component pair are extracted as constraints. These constraints typically originate from the design intent of the garment pattern drawing, where the relative positional parameters under the garment pattern size are considered to be in an ideal state or as a design benchmark.

[0188] Step 4: Based on the relationship type and constraints of the component pair, analyze the relative position parameters under different sizes to identify the assembly coordination of the component pair as the size changes.

[0189] For example, for each component pair, the deviation of each relative position parameter from the constraints under each size is calculated to identify sizes with abnormalities.

[0190] In summary, ensuring that the multi-size patterns after grading are not only numerically reasonable but also spatially coordinated and sewn together is crucial. By quantitatively analyzing the changes in the relative position parameters of component pairs with size variations, potential assembly risks can be intelligently identified, providing more comprehensive quality assurance for custom garment production.

[0191] Furthermore, the relative positional parameters under different sizes are analyzed, including the following steps: Step 1: Compare the relative position parameters of each size with the constraints to obtain the degree of deviation of each relative position parameter under different sizes.

[0192] The degree of deviation can be calculated in ways such as absolute deviation and relative deviation.

[0193] The expression for absolute deviation is as follows:

[0194] in, This is the absolute deviation value. Let i be the value of the i-th parameter in the j-th size. These are constraint values.

[0195] The expression for relative deviation is as follows: , Not 0 in, This represents the relative deviation value.

[0196] Step 2: Retrieve the deviation thresholds corresponding to different relative position parameters from the deviation threshold database.

[0197] The deviation threshold can be set separately for different parameters. For example, the spatial distance threshold can be set to 0.2cm, the overlapping area threshold to 0.15cm, and the angle threshold to 0.5°.

[0198] Step 3: Identify abnormal relative position parameters whose deviation exceeds the corresponding deviation threshold, and mark the sizes with abnormal relative position parameters as abnormal sizes.

[0199] It should be noted that for each relative position parameter, the deviation value for each size is compared with the corresponding deviation threshold. If the deviation exceeds the corresponding deviation threshold, the relative position parameter is considered an abnormal relative position parameter for that size. If any one relative position parameter is abnormal, the entire size is marked as abnormal.

[0200] Step 4: Calculate the proportion of abnormal sizes of the component across all sizes to obtain the size risk value.

[0201] Specifically, the size anomaly ratio is obtained based on the ratio between the number of abnormal sizes and the total number of sizes. The size anomaly ratio is then mapped in the risk mapping table to obtain the corresponding size risk value.

[0202] Step 5: Based on the size risk value, identify the assembly compatibility of the component with size variations.

[0203] Specifically, size risk values ​​can be mapped in a risk level mapping table to obtain the corresponding assembly coordination risk level. An early warning message is generated when the assembly coordination risk level exceeds a preset level.

[0204] It should be noted that different parameters of the same component may deviate to varying degrees, and a comprehensive consideration can be made.

[0205] Specifically, the weights of each relative position parameter are determined based on the relationship type of the component pairs.

[0206] It should be noted that different relationship types have varying sensitivities to relative position parameters, therefore, weights are assigned to different relative position parameters. For example, in a stitching relationship: the overlapping area parameter has the highest weight (affecting sewing feasibility); in a nested relationship: the spatial distance parameter has the highest weight (affecting assembly stability); and in a dependency relationship: the angle deviation parameter has the highest weight (affecting functional implementation). For a stitching relationship: the overlapping area parameter has a weight of 0.5, the spatial distance parameter has a weight of 0.3, and the angle deviation parameter has a weight of 0.2.

[0207] Then, based on the weights of each relative position parameter, the maximum deviation of each relative position parameter across all sizes is weighted and summed to obtain the position parameter risk value.

[0208] Specifically, for each relative positional parameter, the maximum deviation across all sizes is taken, reflecting the maximum potential problem level for that relative positional parameter. Therefore, the positional parameter risk value comprehensively considers the maximum deviation of each relative positional parameter.

[0209] Then, based on the position parameter risk value and the size risk value, the assembly coordination of the component with size variations is identified.

[0210] For example, based on preset weights, the risk values ​​of position parameters and size are weighted and summed to obtain the final risk value. This final risk value is then mapped in a risk level mapping table to obtain the corresponding assembly coordination risk level.

[0211] In summary, by verifying the cross-size process coordination between semantic graphs of different garment structures, potential contradictions between grading rules and between grading rules and pattern structures can be automatically identified. This effectively captures and warns of potential design defects caused by user input errors, contradictory grading rule descriptions, or incompatibility between customer-provided process parameters and pattern structures. Problems can be identified and corrected before production begins, avoiding large-scale garment scrapping due to grading rule errors.

[0212] Based on the same idea, some embodiments of this application also provide devices and non-volatile computer storage media corresponding to the above methods.

[0213] Figure 2 A schematic diagram of a garment process analysis device provided in this application embodiment includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform any of the above-described garment process analysis methods.

[0214] Some embodiments of this application provide a non-volatile computer storage medium for garment process analysis, which stores computer-executable instructions capable of executing any of the above-described garment process analysis methods.

[0215] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.

[0216] The devices and media provided in this application are one-to-one with the methods. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0217] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0218] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0219] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0220] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0221] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0222] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0223] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0224] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0225] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the technical principles of this application should fall within the protection scope of this application.

Claims

1. A method for analyzing garment manufacturing processes, characterized in that, The method includes: From the garment pattern drawings uploaded by the client, multi-level process features reflecting the garment structure are extracted; the multi-level process features include macroscopic structural features reflecting the garment's outline structure, component geometric features, and microscopic process features reflecting process details. Based on the aforementioned multi-level process features, a garment structure semantic graph reflecting the garment structure hierarchy and process relationships is constructed. When the clothing order is customized in a multi-size customization mode, if the clothing order does not have clothing pattern drawings for other sizes, extract the grading process parameter information for other sizes from the clothing order. Based on the grading process parameter information, the relevant node parameters in the garment structure semantic graph are graded and deduced to obtain garment structure semantic graphs for other sizes. The hierarchical features of the garment are extracted from the garment structure semantic map and the garment structure semantic maps of other sizes, respectively, to generate a multi-size process parameter description of the garment. The construction of a garment structure semantic graph reflecting the garment structure hierarchy and process relationships based on the multi-level process features specifically includes: Corresponding graph nodes are established for the macroscopic structural features, component geometric features, and microscopic process features, respectively; the macroscopic structural features correspond to the root node, the component geometric features correspond to the child nodes, and the microscopic process features correspond to the leaf nodes; Based on the technological relationships between the features in the garment pattern drawing, edges are established between the nodes in the drawing; the edges include the membership edges connecting the root node and the child node, the assembly edges connecting the child nodes, the assembly edge relationships including at least one of the sewing relationship, nesting relationship and dependency relationship, and the attachment edges connecting the child node and the leaf node. The parameter information in the multi-level process features is assigned to the corresponding graph nodes as node attributes to obtain the garment structure semantic graph. Based on the grading process parameter information, grading deduction is performed on the relevant node parameters in the garment structure semantic graph to obtain garment structure semantic graphs for other sizes, specifically including: Identify the target node corresponding to the grading process parameter information in the garment structure semantic graph; If the target node is a macroscopic structure node, query the associated component nodes connected to the target node from the clothing structure semantic graph; Based on the geometric relationship between the target node and the associated component node in the garment structure semantic graph, the grading process parameter information is converted into an equivalent displacement of the associated component node. If the target node is a component node or a micro-process node, the grading process parameter information is converted into a direct displacement amount of the component node or the micro-process node. Based on the hierarchical structure of the garment structure semantic graph, the equivalent displacement and / or the direct displacement, the garment structure semantic graphs for other sizes are reconstructed and generated. The method further includes: Extract the first numerical sequence of the same target node under different sizes from the garment structure semantic graph and the garment structure semantic graphs of other sizes; Identify spatially adjacent associated nodes to the target node and extract the second numerical sequence of associated nodes under different sizes; For the first numerical sequence and the second numerical sequence, respectively, a function is fitted with size as the independent variable and parameter value as the dependent variable to obtain a first fitting function and a second fitting function. Analyze the mathematical relationship between the first fitting function and the second fitting function, and calculate the similarity that characterizes the consistency of the relationship between the first fitting function and the second fitting function; The similarity is compared with a preset similarity threshold to determine the parameter coordination between the target node and the associated node as the size changes, thereby completing the verification of the cross-size process coordination of the garment.

2. The method according to claim 1, characterized in that, The step of converting the grading process parameter information into an equivalent displacement of the associated component node based on the geometric relationship between the target node and the associated component node in the garment structure semantic graph specifically includes: Based on the geometric relationship between the target node and the associated component nodes in the garment structure semantic graph, the allocation rule between the size change of the target node and the displacement of each associated component node is determined; According to the allocation rule, the dimensional change in the grading process parameter information is allocated to the equivalent displacement of each associated component node.

3. The method according to claim 1, characterized in that, After comparing the similarity with a preset similarity threshold to determine the parameter consistency between the target node and associated nodes as size changes, the method further includes: Identify component pairs with stitching, nesting, and dependency relationships from the garment structure semantic map; From the garment structure semantic map and the garment structure semantic maps of other sizes, extract the relative position parameters of the component pairs under different sizes; the relative position parameters include at least one of spatial distance parameters, angle deviation parameters, and overlapping area parameters; The relative position parameters of the component pairs in the garment structure semantic diagram are used as constraints. Based on the relationship type and constraints of the component pairs, the relative position parameters under different sizes are analyzed to identify the assembly coordination of the component pairs as the size changes.

4. The method according to claim 1, characterized in that, When the garment pattern drawing is in bitmap format, extract multi-layered technological features reflecting the garment structure from the garment pattern drawing uploaded by the client, specifically including: The garment pattern diagram is decomposed into multiple scales to obtain images at different scale levels. The scale levels include macro scale, meso scale, and micro scale. The resolution of the macro scale is lower than that of the meso scale, the resolution of the meso scale is lower than that of the micro scale, and the resolution of the micro scale is the same as that of the garment pattern diagram. The overall outline features and structural segmentation lines of the garment are extracted from the macro-scale hierarchical image to obtain macro-structural features. The geometric features and relative positional relationships of garment components are extracted from the meso-scale hierarchical image to obtain component geometric features. The process details are extracted from the micro-scale hierarchical image to obtain micro-process features. The process details include at least one of sewing symbols, dart markings, and knife edge markings.

5. The method according to claim 1, characterized in that, The hierarchical features of the garment are extracted from the garment structure semantic map and the garment structure semantic maps of other sizes, respectively, to generate a multi-size process parameter description of the garment, specifically including: According to the hierarchical structure of the garment structure semantic graph, the garment structure semantic graph and the garment structure semantic graphs of other sizes are traversed layer by layer in the order from the root node to the leaf node. In each level of traversal, extract the node identifier, node attributes, and edge relationship information between the current level node and its child nodes; According to the hierarchical structure, the extracted node identifiers, node attributes, and edge relationship information are used as process parameter descriptions for each size.

6. The method according to claim 1, characterized in that, The method further includes: When the clothing order is customized in a single-size customization mode, the hierarchical features of the clothing are extracted from the clothing structure semantic graph to generate a description of the clothing's process parameters.

7. A garment manufacturing process analysis device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform a garment process analysis method according to any one of claims 1-6.

Images