Virtual fitting oriented physical modeling and simulation method for drape and fit of garment

Abstract

The application discloses a virtual fitting-oriented garment drape and fit physical modeling and simulation method, relates to the technical field of garment engineering, and specifically comprises the following steps: S100, garment picture preprocessing; S200, garment information extraction and attribute setting; S300, garment drape physical modeling; S400, garment fit quantitative analysis modeling; S500, drape and fit fusion simulation; and S600, simulation result output and optimization. The application takes garment front and back pictures as input, combines virtual human body model skeleton node features, constructs differentiated drape physical models and quantitative fit analysis models, realizes real simulation of drape effect and objective evaluation of fit degree, simultaneously considers simulation efficiency, adapts to the digital fitting needs of massive garments of garment e-commerce, and provides virtual fitting experience with high reality and high reference for users.

Description

Technical Field

[0001] This invention relates to the fields of computer graphics, clothing engineering and physical simulation technology, and in particular to a physical modeling and simulation method for clothing drape and fit for virtual fitting. Background Technology

[0002] With the development of the apparel e-commerce and digital fashion industries, 3D virtual try-on systems have become an important bridge connecting producers and consumers. Users' demands for the realism and accuracy of try-on experiences continue to rise. Clothing drape and fit are core indicators of virtual try-on. Drape reflects the natural draping and wrinkling characteristics of fabric under the influence of gravity and human body shape, while fit reflects the degree to which clothing conforms to key areas of the body such as shoulders, waist, and hips. The accurate simulation of these two aspects directly determines the reference value of the try-on results.

[0003] Current mainstream virtual try-on methods mostly focus on 3D mesh modeling of clothing and basic physical stitching simulation, which has significant shortcomings in drape and fit simulation: First, traditional fabric physical models do not establish differentiated drape parameter systems for different fabrics such as silk, cotton, linen, and wool, resulting in similar drape effects across different fabrics and significant deviations from reality; Second, fit evaluation relies on subjective human judgment and lacks an objective evaluation system based on human skeleton and clothing contour, failing to accurately reflect fit deviations; Third, the linkage between clothing modeling and virtual human body models is insufficient, failing to coordinate the simulation of human skeletal nodes, posture, and changes in clothing drape and fit, making the simulation prone to distortion when posture changes; Fourth, traditional simulation calculations are inefficient, and complex drape deformation calculations easily cause lag, failing to meet the rapid try-on needs of e-commerce platforms with massive SKUs.

[0004] Meanwhile, the existing technology for setting up sewing information for garment patterns is complex, requires professional garment design knowledge, and involves a large number of two-dimensional patterns, resulting in high modeling costs and long cycles. Hand-drawn or parametric modeling methods are difficult to quickly adapt to the simulation needs of different styles and fabrics, making it difficult to implement in actual e-commerce scenarios.

[0005] Therefore, it is necessary to invent a physical modeling and simulation method for clothing drape and fit in virtual fitting to solve the above problems. Summary of the Invention

[0006] The purpose of this invention is to provide a physical modeling and simulation method for clothing drape and fit in virtual try-on. Using front and back images of clothing as input, and combining the skeletal node features of a virtual human body model, a differentiated physical model of drape and a quantitative analysis model of fit are constructed. This achieves realistic simulation of drape and objective evaluation of fit, while also considering simulation efficiency. It adapts to the digital try-on needs of e-commerce platforms with their massive collections of clothing, providing users with a highly realistic and informative virtual try-on experience, thus solving the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a physical modeling and simulation method for clothing drape and fit in virtual fitting, specifically including the following steps: S100, Clothing Image Preprocessing: Input clear front and back images of clothing without obvious wrinkles, eliminate shooting angle and focal length deviations, and obtain standardized two-dimensional contour data; S200, Clothing Information Extraction and Attribute Setting: Based on the preprocessed contour data, core information is extracted and relevant attributes are set to provide support for subsequent modeling. S300, Physical Modeling of Garment Drape: Based on the normalized contour point set, and combined with fabric properties, an improved spring proton model and a Delaunay triangulated drape physical model are constructed to achieve highly realistic and differentiated simulation. S400, quantitative analysis and modeling of clothing fit: Based on the spatial coordinates of virtual human skeleton nodes and clothing outline, a quantitative evaluation model is constructed to achieve an objective analysis of fit from six key parts: shoulder, chest, waist, hip, sleeve, and pant leg. S500, simulation that blends drape and fit: Design a linkage simulation algorithm to achieve dynamic coordination between sag simulation and fit evaluation, ensuring that the results have both physical realism and fitting accuracy; S600, Simulation Result Output and Optimization: After completing the linkage simulation, the system outputs multi-dimensional fitting results and provides parameter adjustment functions to enhance the interactive experience and reference value.

[0008] Preferably, S100 specifically includes: S101: Use image segmentation algorithm to remove background, extract the outer contour of clothing through contour detection algorithm, retain style features and obtain two-dimensional contour point set; S102: Combines the directional bounding box algorithm to detect shooting angle deviation, rotates and adjusts around the outline centroid point to make the outer outline of the garment upright within the coordinate axis, eliminating the outline deformation caused by tilt; S103: Perform coordinate mapping and normalization on the contour point set, uniformly mapping the coordinates to the 0-1 range, eliminating the difference in contour size caused by focal length, and making the contour data under different shooting conditions comparable.

[0009] Preferably, S200 specifically includes: S201: Utilizes a support vector machine classifier to intelligently classify clothing images, identify clothing categories, and extract outline and basic size features; S202: Extract the skeleton from virtual human models of different body types and postures to obtain the spatial coordinates and actual lengths of 26 skeletal nodes such as wrists, shoulders, waist, and hips; S203: Combining clothing category with the bone length of the corresponding human body part, matching the pattern-specific size scaling factor, setting the basic size of clothing, and ensuring adaptation to the virtual human body model; S204: Based on actual fabric types such as silk, cotton, linen, and denim, basic physical property parameters such as elastic modulus and bending stiffness are set to provide core basis for differentiated simulation of drape.

[0010] Preferably, S300 specifically includes: S301: Convert the front and back contour point sets of the garment into front and back patterns, obtain discrete contour point sets through uniform segmentation, generate a three-dimensional mesh model through Delaunay triangulation, determine the number of triangle moments according to the number of contour points, and adjust the triangle moment density in combination with the fabric surface density. The greater the surface density, the higher the density. S302: Improve the traditional spring-proton model by constructing a three-layer spring structure. The tension spring connects adjacent grid protons to simulate the tensile properties of the fabric, and the stiffness is positively correlated with the elastic modulus of the fabric. The bending spring connects non-adjacent protons in the same direction to simulate the bending and wrinkling properties, and the stiffness is positively correlated with the bending stiffness of the fabric. The gravity spring simulates the effect of gravity on the suspension, and the tension is related to the fabric surface density and the spatial position of the protons. S303: The stiffness of the three-layer springs is differentiated according to the characteristics of the fabric. Lightweight silk is set with low stiffness parameters to simulate soft drape characteristics, while heavy wool and denim are set with high stiffness parameters to simulate crisp and wrinkle-resistant characteristics. Medium-strength fabrics such as cotton and linen are set with intermediate values ​​to restore the real texture. S304: The 3D mesh model is precisely placed in the corresponding spatial position of the virtual human body. Collision detection is used to avoid intersection with the human body model. The initial position of the proton is finely adjusted in combination with the human body skeletal node features to ensure the accurate fit of the basic clothing.

[0011] Preferably, S400 specifically includes: S401: Extract the spatial coordinates of the skeletal nodes of six key parts of the human body, and determine the human body contour boundary of each part based on the node distribution; S402: Extract the outline boundary points of the corresponding parts of the clothing 3D mesh model to ensure that they correspond one-to-one with the outline points of the human body; S403: Calculate the normalized Euclidean distance between the boundary points of the clothing and the corresponding parts of the human body to eliminate the influence of body shape differences and serve as a basic indicator for evaluating fit. S404: Set clear evaluation thresholds and grading standards. Based on the distance value, the fit is divided into five levels: close-fitting, slightly loose, moderate, slightly loose, and too loose. If the clothing outline exceeds the human body outline, it is judged as too tight. Each level is matched with a corresponding score range. S405: Based on the key points of human body clothing, different weights are assigned to six parts, with the shoulders and waist having the highest weight, followed by the chest and hips, and then the sleeves and trouser legs. The scores of each part are weighted and averaged to obtain the overall fit score of the clothing.

[0012] Preferably, S500 specifically includes: S501: Dynamically simulates the physical model of draping on a virtual human body model, simulates the draping deformation of clothing under the combined action of gravity and human body shape, updates the proton space coordinates of the grid in real time, and restores the natural draping and wrinkle state. S502: At each time step of the suspension simulation, the fit analysis model is called in real time to calculate the indicators of each part and the overall score. If the condition of being too tight or too loose is detected, the proton motion constraint is adjusted according to the deformation trend to avoid simulation distortion. S503: It adopts a step-by-step simulation optimization algorithm. First, it performs coarse-grained overall simulation to quickly determine the overall draping shape of the garment and reduce the amount of calculation. Then, it performs fine-grained and refined simulation on highly related parts such as the shoulder and sleeve joint, waist, and hips to restore local details. S504: Adds fabric creep simulation to simulate the slow sagging deformation of parts such as dress hems and shirt hems under long-term gravity, making the draping effect closer to reality.

[0013] Preferably, S600 specifically includes: S601: Outputs a 3D simulation image of the garment's draping effect, clearly showing the wearing effect, draping shape, and local wrinkles. It also outputs the fit score and evaluation conclusion for each key part and the overall fit, providing an intuitive quantitative reference. S602: If the user is not satisfied with the results, they can adjust parameters such as fabric physical properties and garment size through the interface. The system will automatically resimulate and generate multiple sets of results for comparison and selection. S603: Accurately maps the colors, patterns, textures, and fabric details of the original clothing images to a 3D mesh model, uses a Lambertian reflection model to set natural lighting, and maps accessories such as buttons and zippers to enhance the realism of the simulation.

[0014] Compared with the prior art, the present invention has at least the following advantages: 1. The draping simulation is highly realistic. By assigning different parameters to the three-layer spring model and the fabric, the draping characteristics of various fabrics are accurately restored. Gravity and creep simulation are added to make the deformation and wrinkles more realistic. 2. The fit evaluation is objective and quantitative. A quantitative evaluation system is constructed from six key parts, and the overall score is obtained by weighted average, which solves the drawbacks of subjective judgment. 3. Highly efficient modeling and simulation: Using clothing images as input, the system automatically completes contour extraction and modeling. The step-by-step simulation algorithm improves computational efficiency while ensuring accuracy, meeting the real-time interactive needs of e-commerce. 4. The suspension and fit are linked and coordinated. The fit is monitored and adjusted in real time during the simulation to avoid distortion and ensure that the result has both physical properties and fitting accuracy. 5. It has strong scene compatibility, can adapt to virtual human body models of different body types and postures, supports simulation of various clothing styles, and is suitable for multiple scenarios such as e-commerce, design, and offline fitting, making it highly practical. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the system flow of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] This invention provides, for example Figure 1 The physical modeling and simulation method for clothing drape and fit in virtual fitting, as shown, specifically includes the following steps: S100, Clothing Image Preprocessing: Input clear front and back images of clothing without obvious wrinkles. Eliminate shooting angle and focal length deviations to obtain standardized two-dimensional contour data. Specifically, this includes: using image segmentation algorithms to remove the background; extracting the outer contour of the clothing using contour detection algorithms, preserving style features and obtaining a two-dimensional contour point set; combining directional bounding box algorithms to detect shooting angle deviations, rotating and adjusting around the contour centroid to make the outer contour of the clothing upright within the coordinate axes, eliminating contour deformation caused by tilt; and performing coordinate mapping and normalization on the contour point set, uniformly mapping the coordinates to the 0-1 range to eliminate contour size differences caused by focal length, making contour data from different shooting conditions comparable. S200, Clothing Information Extraction and Attribute Setting: Based on the preprocessed contour data, core information is extracted and relevant attributes are set to support subsequent modeling. Specifically, this includes: using a support vector machine classifier to intelligently classify clothing images, identify clothing categories, and extract contour and basic size features; extracting skeletons from virtual human models of different body types and postures to obtain the spatial coordinates and actual lengths of 26 skeletal nodes such as wrists, shoulders, waist, and hips; matching the clothing category with the corresponding skeletal lengths of the human body, matching the pattern-specific size scaling factor, and setting the basic size of the clothing to ensure compatibility with the virtual human model; and setting basic physical property parameters such as elastic modulus and bending stiffness according to actual fabric types such as silk, cotton, linen, and denim to provide core basis for differentiated simulation of drape. S300, Physical Modeling of Garment Drape: Based on normalized contour point sets and combined with fabric properties, an improved spring-proton model and a Delaunay triangulation-based drape physical model are constructed to achieve highly realistic and differentiated simulations. Specifically, this includes: converting the front and back contour point sets of the garment into front and back patterns; obtaining discrete contour point sets through uniform segmentation; generating a 3D mesh model through Delaunay triangulation; determining the number of triangular moments based on the number of contour points; and adjusting the triangular moment density based on the fabric's surface density—a higher surface density results in a higher overall density. The traditional spring-proton model is improved by constructing a three-layer spring structure, with tension springs connecting adjacent mesh protons to simulate fabric stretching characteristics, where stiffness is positively correlated with the fabric's elastic modulus; and bending springs connecting the same type of springs. For non-adjacent protons, the bending and wrinkling characteristics are simulated, and the stiffness is positively correlated with the bending stiffness of the fabric. Gravity springs simulate the effect of gravity on drape, and the tension is related to the fabric surface density and the spatial position of the protons. The stiffness of the three layers of springs is differentiated according to the fabric characteristics: light silk is assigned low stiffness parameters to simulate soft drape characteristics, heavy wool and denim are assigned high stiffness parameters to simulate crisp and wrinkle-resistant characteristics, and medium-strength fabrics such as cotton and linen are assigned intermediate values ​​to restore realistic texture. The three-dimensional mesh model is accurately placed in the corresponding spatial position of the virtual human body. Collision detection is used to avoid intersection with the human body model, and the initial position of the protons is finely adjusted in combination with the human skeleton node features to ensure accurate basic fit of the clothing. S400, quantitative analysis and modeling of clothing fit: Based on the spatial coordinates of virtual human skeleton nodes and clothing outlines, a quantitative evaluation model is constructed to objectively analyze the fit of clothing from six key areas: shoulders, chest, waist, hips, sleeves, and trouser legs. Specifically, this includes: extracting the spatial coordinates of the skeleton nodes of the six key human body areas; determining the human body outline boundaries for each area based on node distribution; extracting the outline boundary points of the corresponding areas of the clothing's 3D mesh model to ensure a one-to-one correspondence with the human body outline points; calculating the normalized Euclidean distance between the clothing and the corresponding boundary points of the human body to eliminate the influence of body shape differences, serving as the basic indicator for fit evaluation; setting clear evaluation thresholds and grade standards, classifying fit into five levels—fitting, slightly loose, moderate, slightly loose, and too loose—based on distance values; clothing outlines exceeding the human body outline are judged as too tight, with each level matched to a corresponding score range; and assigning differentiated weights to the six areas based on key considerations for human attire, with shoulders and waist receiving the highest weight, followed by chest and hips, and then sleeves and trouser legs. The scores for each area are then weighted and averaged to obtain the overall fit score of the clothing. S500, simulation that blends drape and fit: A linked simulation algorithm was designed to achieve dynamic coordination between drape simulation and fit evaluation, ensuring that the results are both physically realistic and accurately tailored. Specifically, this includes: dynamically simulating the drape physical model on a virtual human body model to simulate the drape deformation of clothing under the combined effects of gravity and human body shape, updating the proton space coordinates of the mesh in real time to restore the natural drape and wrinkle state; at each time step of the drape simulation, calling the fit analysis model in real time to calculate the indicators of each part and the overall score; if excessive tightness or looseness is detected, adjusting the proton motion constraints according to the deformation trend to avoid simulation distortion; employing a step-by-step simulation optimization algorithm, first performing a coarse-grained overall simulation to quickly determine the overall drape shape of the clothing and reduce computational load, then performing fine-grained and refined simulations of highly correlated parts such as the shoulder and sleeve joints, waist, and hips to restore local details; and incorporating fabric creep characteristic simulation to simulate the slow drape deformation of parts such as dress hems and shirt hems under long-term gravity, making the drape effect closer to reality. S600, Simulation Result Output and Optimization: After completing the linkage simulation, the system outputs multi-dimensional fitting results and provides parameter adjustment functions to enhance the interactive experience and reference value. Specifically, it outputs a 3D simulation image of the garment's drape effect, clearly showing the wearing effect, drape shape, and local wrinkles. It also outputs scores and evaluation conclusions for the fit of key parts and the overall fit, providing intuitive quantitative references. If the user is not satisfied with the results, they can adjust parameters such as fabric physical properties and garment size through the interface. The system will automatically re-simulate and generate multiple sets of results for comparison and selection. The system accurately maps the colors, patterns, textures, and fabric textures of the original garment images to a 3D mesh model, uses a Lambertian reflection model to set natural lighting, and maps accessories such as buttons and zippers to enhance the visual realism of the simulation.

[0018] Taking the virtual try-on of a women's silk dress as an example, the implementation process of this method is fully explained: (1) Experimental preparation: Input high-resolution images of the front and back of a sleeveless silk dress without wrinkles and taken horizontally from the front; select a female virtual human body model with a standard body size of 165cm / 50kg standing naturally, extract the coordinates of 26 skeletal nodes, and focus on obtaining the bone length and contour boundary of the shoulder, waist, hip and other parts; set the basic physical parameters of low elastic modulus and low bending stiffness of silk.

[0019] (2) Image preprocessing: The background was removed by the U-Net algorithm, and the Canny algorithm was used to extract a two-dimensional point set containing 864 contour points. A 3° shooting angle deviation was detected, and the deformation was eliminated after rotation adjustment. The contour point set was normalized to obtain standardized data.

[0020] (3) Information extraction and attribute setting: The support vector machine classifier determines that it is a sleeveless silk dress and extracts the basic size features; extracts the bone length of key parts of the human body, matches the dress pattern scaling coefficient, and sets the basic size; matches the three-layer spring low stiffness parameter for the silk to adapt to the soft draping characteristics.

[0021] (4) Physical modeling of drape: The front and back patterns of the dress are triangulated by Delaunay to generate a three-dimensional mesh model with approximately 43,200 triangular moments; a three-layer spring proton model is built and the silk parameters are assigned; the model is placed in the corresponding position of the virtual human body, and the initial position of the proton is finely adjusted to ensure that the shoulder and waist parts fit accurately.

[0022] (5) Fit Quantification Analysis: 200 contour boundary points of five key parts of the human body, namely the shoulder, chest, waist, hip and skirt, were extracted, and 200 boundary points of the corresponding parts of the clothing were extracted simultaneously; the normalized Euclidean distance was calculated, and the results showed that the waist and hip were well fitted, the shoulder and chest were slightly loose, and the skirt was moderate; according to the standard matching score, the waist and hip were 95 points, the shoulder and chest were 85 points, and the skirt was 75 points. After weighted average, the overall fit was 87.2 points, and the evaluation was "slightly loose, and the overall fit was good".

[0023] (6) Fusion simulation: Dynamic simulation of the dress restores the natural drape of silk and the soft folds of the skirt; real-time monitoring of fit within 10 time steps. Due to the improved drape and fit of the skirt, the score rises to 78 points, and the overall score reaches 88 points; step-by-step simulation is adopted, with overall coarse-grained simulation plus fine-grained simulation of the skirt and waist, with a total time of 0.8s, which meets the real-time requirements; creep simulation is added, and the skirt will sag slightly and slowly in the later stage, which is close to the real state.

[0024] (7) Output results: Output a highly realistic 3D simulation image of the draping effect, with natural texture and lighting; Output a fit evaluation report, including scores for each part, an overall score of 88, and evaluation conclusions; Provide options for adjusting fabric and size parameters, and support users to re-simulate and compare.

[0025] In summary, this invention, in apparel e-commerce platforms, can establish digital simulation models for massive amounts of clothing. Users can upload their body data to achieve personalized try-ons, obtaining realistic drape effects and quantitative fit evaluations, solving the pain points of "unseen and ill-fitting" in online shopping, increasing conversion rates and reducing return rates. In apparel digital design platforms, enterprises can quickly simulate the drape and fit of different fabrics and styles during the design stage, reducing physical sampling costs, shortening R&D cycles, and promoting the digital transformation of design. In offline smart fitting devices, combining body sensing and virtual human body modeling technologies allows users to experience the try-on effect without actually putting on and taking off the clothes, improving store service efficiency and experience. In the field of customized clothing, it can accurately simulate the drape and fit of customized clothing based on customer body data, confirming design schemes in advance, improving customization accuracy and customer satisfaction. It solves the core problems of current virtual try-ons, such as drape distortion, vague fit evaluation, and low simulation efficiency, achieving realism, quantification, and efficiency in virtual try-ons, promoting the transformation of the apparel industry from traditional models to digital, online-offline integrated models, and has broad application prospects.

[0026] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A physical modeling and simulation method for clothing drape and fit in virtual fitting, characterized in that, Specifically, the following steps are included: S100, Clothing Image Preprocessing: Input clear front and back images of clothing without obvious wrinkles, eliminate shooting angle and focal length deviations, and obtain standardized two-dimensional contour data; S200, Clothing Information Extraction and Attribute Setting: Based on the preprocessed contour data, core information is extracted and relevant attributes are set to provide support for subsequent modeling. S300, Physical Modeling of Garment Drape: Based on the normalized contour point set, and combined with fabric properties, an improved spring proton model and a Delaunay triangulated drape physical model are constructed to achieve highly realistic and differentiated simulation. S400, quantitative analysis and modeling of clothing fit: Based on the spatial coordinates of virtual human skeleton nodes and clothing outline, a quantitative evaluation model is constructed to achieve an objective analysis of fit from six key parts: shoulder, chest, waist, hip, sleeve, and pant leg. S500, simulation that blends drape and fit: Design a linkage simulation algorithm to achieve dynamic coordination between sag simulation and fit evaluation, ensuring that the results have both physical realism and fitting accuracy. S600, Simulation Result Output and Optimization: After completing the linkage simulation, the system outputs multi-dimensional fitting results and provides parameter adjustment functions to enhance the interactive experience and reference value.

2. The physical modeling and simulation method for clothing drape and fit in virtual fitting as described in claim 1, characterized in that: Specifically, S100 includes: S101: Use image segmentation algorithm to remove background, extract the outer contour of clothing through contour detection algorithm, retain style features and obtain two-dimensional contour point set; S102: Combines the direction bounding box algorithm to detect shooting angle deviation, rotates and adjusts around the outline centroid point to make the outer outline of the garment upright within the coordinate axis, eliminating the outline deformation caused by tilt; S103: Perform coordinate mapping and normalization on the contour point set, uniformly mapping the coordinates to the 0-1 range, eliminating the difference in contour size caused by focal length, and making the contour data under different shooting conditions comparable.

3. The physical modeling and simulation method for clothing drape and fit in virtual fitting as described in claim 2, characterized in that: Specifically, S200 includes: S201: Utilizes a support vector machine classifier to intelligently classify clothing images, identify clothing categories, and extract outline and basic size features; S202: Extract the skeleton from virtual human models of different body types and postures to obtain the spatial coordinates and actual lengths of 26 skeletal nodes such as wrists, shoulders, waist, and hips; S203: Combining clothing category with the bone length of the corresponding human body part, matching the pattern-specific size scaling factor, setting the basic size of clothing, and ensuring adaptation to the virtual human body model; S204: Based on actual fabric types such as silk, cotton, linen, and denim, basic physical property parameters such as elastic modulus and bending stiffness are set to provide core basis for differentiated simulation of drape.

4. The physical modeling and simulation method for clothing drape and fit in virtual fitting as described in claim 3, characterized in that: Specifically, S300 includes: S301: Convert the front and back contour point sets of the garment into front and back patterns, obtain discrete contour point sets through uniform segmentation, generate a three-dimensional mesh model through Delaunay triangulation, determine the number of triangle moments according to the number of contour points, and adjust the triangle moment density in combination with the fabric surface density. The greater the surface density, the higher the density. S302: Improve the traditional spring-proton model by constructing a three-layer spring structure. The tension spring connects adjacent grid protons to simulate the tensile properties of the fabric, and the stiffness is positively correlated with the elastic modulus of the fabric. The bending spring connects non-adjacent protons in the same direction to simulate the bending and wrinkling properties, and the stiffness is positively correlated with the bending stiffness of the fabric. The gravity spring simulates the effect of gravity on the suspension, and the tension is related to the fabric surface density and the spatial position of the protons. S303: The stiffness of the three-layer springs is differentiated according to the characteristics of the fabric. Lightweight silk is set with low stiffness parameters to simulate soft drape characteristics, while heavy wool and denim are set with high stiffness parameters to simulate crisp and wrinkle-resistant characteristics. Medium-strength fabrics such as cotton and linen are set with intermediate values ​​to restore the real texture. S304: The 3D mesh model is precisely placed in the corresponding spatial position of the virtual human body. Collision detection is used to avoid intersection with the human body model. The initial position of the proton is finely adjusted in combination with the human body skeletal node features to ensure the accurate fit of the basic clothing.

5. The physical modeling and simulation method for clothing drape and fit in virtual fitting as described in claim 4, characterized in that: Specifically, S400 includes: S401: Extract the spatial coordinates of the skeletal nodes of six key parts of the human body, and determine the human body contour boundary of each part based on the node distribution; S402: Extract the outline boundary points of the corresponding parts of the clothing 3D mesh model to ensure that they correspond one-to-one with the outline points of the human body; S403: Calculate the normalized Euclidean distance between the boundary points of the clothing and the corresponding parts of the human body to eliminate the influence of body shape differences and serve as a basic indicator for evaluating fit. S404: Set clear evaluation thresholds and grading standards. Based on the distance value, the fit is divided into five levels: close-fitting, slightly loose, moderate, slightly loose, and too loose. If the clothing outline exceeds the human body outline, it is judged as too tight. Each level is matched with a corresponding score range. S405: Based on the key points of human body clothing, different weights are assigned to six parts, with the shoulders and waist having the highest weight, followed by the chest and hips, and then the sleeves and trouser legs. The scores of each part are weighted and averaged to obtain the overall fit score of the clothing.

6. The physical modeling and simulation method for clothing drape and fit for virtual fitting as described in claim 5, characterized in that: The S500 specifically includes: S501: Dynamically simulates the physical model of draping on a virtual human body model, simulates the draping deformation of clothing under the combined action of gravity and human body shape, updates the proton space coordinates of the grid in real time, and restores the natural draping and wrinkle state. S502: At each time step of the suspension simulation, the fit analysis model is called in real time to calculate the indicators of each part and the overall score. If the condition of being too tight or too loose is detected, the proton motion constraint is adjusted according to the deformation trend to avoid simulation distortion. S503: It adopts a step-by-step simulation optimization algorithm. First, it performs coarse-grained overall simulation to quickly determine the overall draping shape of the garment and reduce the amount of calculation. Then, it performs fine-grained and refined simulation on highly related parts such as the shoulder and sleeve joint, waist, and hips to restore local details. S504: Adds fabric creep simulation to simulate the slow sagging deformation of parts such as dress hems and shirt hems under long-term gravity, making the draping effect closer to reality.

7. The physical modeling and simulation method for clothing drape and fit in virtual fitting as described in claim 6, characterized in that: The S600 specifically includes: S601: Outputs a 3D simulation image of the garment's draping effect, clearly showing the wearing effect, draping shape, and local wrinkles. It also outputs the fit score and evaluation conclusion for each key part and the overall fit, providing an intuitive quantitative reference. S602: If the user is not satisfied with the results, they can adjust parameters such as fabric physical properties and garment size through the interface. The system will automatically resimulate and generate multiple sets of results for comparison and selection. S603: Accurately maps the colors, patterns, textures, and fabric details of the original clothing images to a 3D mesh model, uses a Lambertian reflection model to set natural lighting, and maps accessories such as buttons and zippers to enhance the realism of the simulation.

Images