Intelligent display system and method for costume design

Abstract

The invention relates to the technical field of costume design, in particular to an intelligent display system and method for costume design, and the method comprises the following steps: receiving a costume design sketch, employing a convolutional neural network to extract contour features, and generating a structured 3D costume model containing a layout topological relation and fabric physical parameters; obtaining user body type scanning data, and constructing a parameterized virtual human body with an anatomical level and soft tissue dynamic response capability; dynamic fitting of the garment model and a virtual human body is achieved through a physical simulation engine, and garment deformation behaviors changing along with actions are calculated; and analyzing the scene semantic tag input by the user, matching illumination, accessories and background elements in a preset material library, and generating an interactive 3D display scene with physical interaction and light and shadow effects. According to the method, the costume design modeling efficiency and simulation precision can be improved, intelligent matching of the costume style and scene semantics is realized, and the immersion of costume display and the user experience are enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of clothing design, and particularly relates to an intelligent display system and method for clothing design. BACKGROUND

[0002] In the modern clothing design and display process, three-dimensional visualization technology is gradually replacing traditional two-dimensional drawings and physical sample clothing means, and becomes an important bridge connecting creative ideas and user experience. With the development of computer graphics, physical simulation and human-computer interaction technology, automatic modeling of clothing design sketches, personalized fitting of human body and immersive three-dimensional display have become the main direction of industry development. Especially under the promotion of e-commerce platforms and virtual fitting applications, designers and consumers have increasingly high requirements for the reality, interactivity and adaptability of clothing digital presentation effect, which puts higher challenges on modeling accuracy, dynamic simulation capability and scene construction logic.

[0003] However, the prior art still has many limitations in actual application. On the one hand, the traditional three-dimensional clothing generation method based on template matching or manual modeling is difficult to handle complex contour changes and structural details in design sketches, and cannot automatically construct a structured model with physical properties; on the other hand, the existing virtual fitting system mostly uses a general human body model, lacks precise response to individual body shape characteristics and motion driving of users, and is difficult to truly restore the dynamic fitting state of clothing. In addition, scene display mostly relies on static backgrounds and manual matching, lacks intelligent matching mechanism for semantic labels and clothing styles, resulting in insufficient immersion and compatibility of the final display effect, which limits its popularization and application in personalized design display and high-fidelity fitting experience. SUMMARY

[0004] The present application provides an intelligent display system and method for clothing design.

[0005] An intelligent display method for clothing design, comprising the following steps: S1: receiving a clothing design sketch, extracting contour features of the clothing design sketch through a convolutional neural network, generating a structured 3D clothing model, and the structured 3D clothing model comprising a pattern topological relationship and a fabric physical parameter; S2: obtaining user body shape scanning data, generating a parameterized virtual human body based on biomechanical characteristics, wherein the range of motion of the joints of the skeleton is associated with a soft tissue deformation coefficient; S3: mapping the structured 3D clothing model to the surface of the parameterized virtual human body, calculating a dynamic fitting displacement field of clothing vertices and human body grids through a real-time physical simulation engine, and driving the clothing to deform with the human body motion; S4: Analyzing the scene semantic label input by the user, matching a combination of matching elements compatible with the current clothing style and physical parameters from a material library, and generating an interactive 3D display scene including lighting, accessories and background.

[0006] Optionally, the S1 comprises: S11: Receiving a user inputted clothing design sketch, and performing grayscale normalization and noise filtering preprocessing on the clothing design sketch; S12: Inputting the preprocessed clothing design sketch into a pre-trained convolutional neural network, and extracting multi-scale fused clothing design sketch contour features through the encoder-decoder structure of the convolutional neural network; S13: Based on the clothing design sketch contour features, generating an initial framework of a structured 3D clothing model including NURBS surface control points by using a parametric surface reconstruction algorithm.

[0007] Optionally, the S1 further comprises: S14: Performing pattern segmentation processing on the initial framework of the structured 3D clothing model, and establishing pattern topological relationships between patterns through a geometric constraint solver, the pattern topological relationships including sewing connection relationships and overlapping constraint relationships; S15: Matching fabric physical parameters associated with the clothing design sketch contour features from a pre-set fabric database, the fabric physical parameters including bending stiffness coefficient, tensile elastic modulus and friction coefficient; S16: Binding the pattern topological relationships and fabric physical parameters to the initial framework of the structured 3D clothing model, and generating a structured 3D clothing model with physical attributes.

[0008] Optionally, the S2 comprises: S21: Obtaining user body shape scanning data through a three-dimensional scanning device, and performing point cloud registration and hole repair processing on the user body shape scanning data; S22: Extracting biomechanical features from the processed user body shape scanning data, the biomechanical features including bone point position markers, joint rotation axis vectors and body surface muscle distribution heat maps; S23: Driven by the biomechanical features, deforming a parametric human template to generate a parametric virtual human basic mesh with an anatomical level structure, and calculating soft tissue deformation coefficients at main joints.

[0009] Optionally, the S2 further comprises: S24: Establishing a dynamic correlation model of the range of motion of the skeletal joints and the soft tissue deformation coefficients, wherein the range of motion of the skeletal joints is defined by a rotation angle threshold matrix, and the soft tissue deformation coefficients are updated in real time by interpolation according to the joint rotation angle; S25: encode the dynamic correlation model to the vertex weight map of the parameterized virtual human body base mesh, generating a parameterized virtual human body capable of driving soft tissue deformation.

[0010] Optionally, the S3 comprises: S31: spatially align the structured 3D garment model with the parameterized virtual human body, making the initial position of the structured 3D garment model fit the surface of the parameterized virtual human body through a rigid transformation matrix; S32: based on the pattern topological relationship and the fabric physical parameters, establish a collision constraint relationship between the garment vertices and the human body mesh surface in the physical simulation engine, the collision constraint relationship includes a sliding friction threshold and a penetration depth tolerance; S33: drive the parameterized virtual human body to perform a target action, and calculate the surface deformation of the parameterized virtual human body caused by the motion of the skeletal joints in real time through the physical simulation engine.

[0011] Optionally, the S3 further comprises: S34: according to the surface deformation and the collision constraint relationship, solve the dynamic fitting displacement field of the garment vertices relative to the surface of the parameterized virtual human body frame by frame in the physical simulation engine, the dynamic fitting displacement field includes vertex normal displacement component and tangential sliding component; S35: apply the dynamic fitting displacement field to the garment vertices of the structured 3D garment model, drive the structured 3D garment model to produce physical deformation with the action of the parameterized virtual human body.

[0012] Optionally, the S4 comprises: S41: analyze the scene semantic label input by the user, and decompose the scene semantic label into three dimensions of environmental lighting demand, accessory function demand and background space demand; S42: extract the garment style feature vector and the physical parameter constraint set based on the structured 3D garment model, the garment style feature vector includes color distribution histogram and contour geometry style code, and the physical parameter constraint set includes fabric light transmittance threshold and accessory load limit; S43: perform multi-level matching retrieval in the preset material library, including: First level: match the lighting parameters according to the environmental lighting demand, the lighting parameters include color temperature, intensity attenuation curve and shadow definition; Second level: screen accessory models according to the accessory function demand and the physical parameter constraint set, to ensure that the accessory model hanging point load value is less than the accessory load limit in the garment physical parameter constraint set; Third level: retrieve background models according to the background space demand and the garment style feature vector, requiring that the color distribution of the background model and the garment color distribution histogram meet the visual compatibility criteria.

[0013] Optionally, S4 further comprises: S44: combining the matched lighting parameters, accessory models and background models into a combination of collocation elements, and binding the accessory models to the corresponding hanging points of the structured 3D garment model through spatial coordinate transformation; S45: integrating the structured 3D garment model, the parameterized virtual human body and the combination of collocation elements in a physical simulation engine to generate an interactive 3D display scene with dynamic light and shadow effects and physical interaction.

[0014] An intelligent display system for garment design is used to implement the above-mentioned intelligent display method for garment design, comprising the following modules: Sketch processing module: used to receive user inputted garment design sketches, and extract contour features through convolutional neural network to generate a structured 3D garment model including plate topological relationship and fabric physical parameters; Virtual human body modeling module: used to obtain user body scan data, and construct a parameterized virtual human body based on biomechanical features, the parameterized virtual human body having a dynamic correlation model of bone joint range of motion and soft tissue deformation coefficient; Garment fitting simulation module: used to map the structured 3D garment model to the surface of the parameterized virtual human body, and establish collision constraint relationship in a physical simulation engine to realize dynamic fitting and physical deformation of the garment caused by human body action; Scene generation module: used to analyze user inputted scene semantic labels, and match lighting parameters, accessory models and background models in a material library based on garment style features and physical parameter constraints to generate an interactive 3D display scene.

[0015] The present application has the following advantages: The present application automatically extracts multi-level garment sketch contour features through pre-processing and encoding-decoding structure of convolutional neural network, and realizes high-fidelity modeling from two-dimensional sketch to three-dimensional garment model with structure and physical properties by combining NURBS surface reconstruction and plate topological solution. This process does not depend on manual modeling experience, significantly reduces the modeling threshold, improves the three-dimensional modeling efficiency and modeling accuracy, and is especially suitable for non-technical background designers to quickly verify the feasibility of garment structure.

[0016] The present application constructs a parameterized virtual human body model based on scan data, and dynamically models the bone joint range of motion and the soft tissue deformation coefficient, calculates the garment and human body fitting displacement field through a real-time simulation engine, realizes the natural wrinkles, stretching and sliding response of the garment caused by action. Compared with the traditional static display method, this scheme significantly improves the authenticity and dynamic behavior consistency of the wearing effect in the simulation level, enhances the user's perception and judgment of the motion adaptability of the garment.

[0017] The present application, by analyzing the scene semantic label and disassembling it into lighting, accessories and background requirements, while combining the color style and physical parameters of the garment, performs hierarchical matching retrieval and spatial integration, realizes semantic coordination and visual fusion between the garment and the scene. The finally generated interactive 3D display scene has dynamic light and shadow effect and physical interaction response, greatly improving the immersion and dissemination of the garment design proposal in e-commerce display, customer fitting experience and creative aesthetic evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present application or the prior art, below will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only illustrate the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0019] Figure 1 The method flowchart of the embodiment of the present application is shown in the figure. Figure 2 The system flowchart of the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0020] The present application will be described in detail below in combination with the drawings and specific embodiments. It should be noted that in order to make the embodiments more detailed, the following embodiments are the best, preferred embodiments, and other alternative ways can also be used by those skilled in the art to implement them; and the drawings are only used to describe the embodiments more specifically, and are not intended to specifically limit the present application.

[0021] It should be noted that in the specification, "one embodiment", "embodiment", "exemplary embodiment", "some embodiments" and the like indicate that the described embodiments can include specific features, structures or characteristics, but not necessarily every embodiment includes the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in combination with an embodiment, it should be within the knowledge of those skilled in the related art to realize this feature, structure or characteristic in combination with other embodiments (whether or not explicitly described).

[0022] Generally, the terms can be understood at least in part from the use in the context. For example, depending at least in part on the context, the term "one or more" used herein can be used to describe any feature, structure or characteristic in the singular or can be used to describe combinations of features, structures or characteristics in the plural. In addition, the term "based on" can be understood as not necessarily intended to convey a set of exclusive factors, but can instead, depending at least in part on the context, allow the presence of other factors not necessarily explicitly described.

[0023] As Figure 1 shown, an intelligent display method for clothing design includes the following steps: S1: receiving a clothing design sketch, extracting clothing design sketch contour features through a convolutional neural network, generating a structured 3D clothing model, the structured 3D clothing model including pattern topological relationships and fabric physical parameters; S2: obtaining user body scan data, generating a parameterized virtual human body based on biomechanical features, wherein the range of motion of the skeletal joints is associated with soft tissue deformation coefficients; S3: mapping the structured 3D clothing model to the surface of the parameterized virtual human body, calculating the dynamic fitting displacement field of the clothing vertices and the human body mesh through a real-time physics simulation engine, and driving the clothing to deform with the human body motion; S4: analyzing user input scene semantic labels, matching compatible combination of collocation elements from the material library according to the current clothing style and physical parameters, and generating an interactive 3D display scene including lighting, accessories and background.

[0024] S1 includes: S11, clothing design sketch preprocessing: receiving user input clothing design sketch image data, using grayscale normalization method to standardize the original image to eliminate the interference of light changes on feature extraction. The normalization process uses the following formula: ; wherein, represents the pixel value of the original image, and represent the maximum and minimum values of the image pixel value respectively.

[0025] Then, a Gaussian filter is used to remove noise, with a filter kernel size of , and a variance parameter to ensure that the sketch contour edge is smooth without distortion.

[0026] S12, convolutional neural network multi-scale contour feature extraction: input the preprocessed image into the pre-trained convolutional neural network, which uses an encoder-decoder structure, the encoder part is composed of multiple layers of convolution and maximum pooling, used to extract low-level to high-level image features; the decoder part uses an upsampling and convolution combined way to restore the spatial resolution.

[0027] The output of the convolutional neural network is a sketch contour feature map fused with multiple scales, denoted as , wherein , are the height and width of the image respectively, is the number of channels.

[0028] To improve contour accuracy, the final output layer of the network uses the Sigmoid activation function to output a probability mask map, which serves as a shape guide for subsequent structure reconstruction.

[0029] S13, Construction of the initial framework for a structured 3D clothing model: based on extracted contour feature maps The surface of the garment is reconstructed using a parametric surface reconstruction algorithm. An initial model framework is generated using a non-uniform rational B-spline modeling method, and its surface is defined as follows: ; in, As control points, These are the weighting coefficients. , These are p-order and q-order B-spline basis functions, respectively. The control point positions are estimated based on the principal axis distribution of the sketch outline and key corner points; a typical example is the feature point set formed by the sleeve area and the shoulder corner.

[0030] S14, Establishing Plate Topology: Plate segmentation based on boundary line clustering is performed on the initial curved surface framework. Several independent plate regions are divided by analyzing the local curvature changes of the NURBS surface and the linear cutting path. Subsequently, the plate topology is constructed using a geometric constraint solver. Sewing connection relationship: indicates the physical connection between two patterns at the edge nodes; Overlapping constraints: Define the stacking order or interleaving relationship to avoid cross-penetration during simulation; Topological relation data in graph structure This indicates that the node Indicates the different versions of the film, edge This indicates physical connections and constraints.

[0031] S15, Fabric Physical Parameter Matching: From the preset fabric database, match the most relevant fabric type based on the sketch outline features (such as curvature distribution, edge density, and side length ratio), and extract its corresponding set of physical attribute parameters. This includes: Bending stiffness coefficient : Indicates the garment's resistance to bending; Tensile modulus of elasticity : Indicates the stretchability of clothing fabric under stress; coefficient of friction Controlling the slippage of clothing on the surface in contact with the human body; For example, if the sketch features numerous draped skirt hems and pleated outlines, it will automatically match a soft silk fabric, with typical parameters being... ≈0.02, ≈300MPa ≈0.1; S16, Structured Model Binding Generation: This involves generating the above-mentioned board topology relationships and physical attribute parameters. Bind to the initial framework of the structured 3D clothing model. This binding process is implemented using node attribute extension, attaching a structure field to each mesh node: ,in, For node coordinates, It is the normal vector. This refers to the section number to which it belongs.

[0032] After binding is completed, a structured 3D clothing model with complete structural definition and physical properties is output, providing a highly controllable model input for subsequent simulation and display stages.

[0033] S2 includes: S21, User Body Shape Scanning Data Processing: First, raw 3D point cloud data of the user's body shape is collected using a 3D scanning device (such as a structured light scanner or a ToF depth camera). ; Multi-view registration is performed on the raw point cloud data, using an iterative nearest-point algorithm to align the scan views, as shown below: ; in, For source point cloud points, For the target point cloud corresponding points, Let be a rotation matrix. It is a translation vector; After registration, the Poisson surface reconstruction algorithm is applied to fill the holes in missing areas (such as the armpits and inner thighs), outputting a complete and continuous user body surface model. ; S22, Biomechanical Feature Extraction: From User Body Shape Surface Model The automatic extraction of biomechanical features mainly includes the following three dimensions: Bone point location marking: Identifies key bone points in the human body (such as the acromion, patella, greater trochanter, etc.) and outputs a set of bone points. Automatic positioning is achieved through morphological template matching and geometric projection; Joint rotation axis vector: for each bone point pair The direction and length of the line vector connecting the two points are calculated and used as the rotation axis of the corresponding joint; Muscle distribution heatmap: Using curvature field analysis and skin reflectance mapping technology, a two-dimensional heatmap describing the muscle thickness and tension state in different areas of the body surface is generated, denoted as... This serves as the basis for subsequent deformation weight adjustments.

[0034] S23, Parametric Virtual Human Body Base Mesh Construction: The extracted biomechanical features are input into the human body template model (such as SMPL or a self-developed mesh), driving its skeleton and surface nodes to deform synchronously, generating a parametric virtual human body base mesh adapted to the user's body shape. The mesh structure is defined as follows: ; in, For the set of vertices, The edge sets form a human body triangular mesh structure. This involves controlling the structure at the skeletal level. During this process, the soft tissue deformation coefficients at each major joint are simultaneously calculated. , representing the degree of coordinated deformation of the skin, muscles, and bones during joint movement, is defined as: ; in, This corresponds to the distance the soft tissue at the joint moves. This represents the change in the rotation angle of the joint.

[0035] S24, Establishment of dynamic correlation model: Construct a dynamic correlation model between the range of motion of skeletal joints and the deformation coefficient of soft tissue to express the mapping relationship between motion and deformation.

[0036] First, define the joint rotation angle threshold matrix, represented as follows: ; Each joint This corresponds to an activity range.

[0037] Then, linear interpolation or spline interpolation is used to update the soft tissue deformation coefficients in real time: ; in, The initial deformation coefficient, The default attitude angle. This is the deformation sensitivity coefficient. This model ensures the naturalness and physical plausibility of the human body's response in dynamic motion simulation.

[0038] S25, Generating a Virtual Human Body Capable of Driving Soft Tissue Deformation: Finally, the dynamically associated model is encoded into the human body mesh structure, and motion-driven binding from bones to surface nodes is achieved through a skinweight map method. Each mesh vertex... Corresponding to a weight vector: ; in, Indicates the first The vertex is affected by the first The degree of influence on each joint ∈1.2... , This represents the total number of joints.

[0039] The generated parametric virtual human body has anatomical structure hierarchy and soft tissue dynamic response capabilities. In the subsequent physical simulation stage, the human body mesh can be directly driven to deform through skeletal movements, so as to achieve a clothing display effect that is consistent with the actual wearer's dynamics.

[0040] S3 includes: S31, Spatial Alignment and Rigid Body Registration: Initial alignment of the structured 3D clothing model with the parametric virtual human body is performed using a rigid body transformation matrix. Achieve spatial coordinate registration: ; in, Represents the rotation matrix. It is a translation vector; The initial registration parameters are automatically estimated by minimizing the difference between the centroids and principal axes of the model's bounding boxes. The transformation is applied to all vertices of the clothing model. , to satisfy: ; This step ensures that the structured 3D clothing model fits the parametric virtual human body surface reasonably in a static state, providing initial boundary conditions for subsequent simulations.

[0041] S32, Establishing Collision Constraints Between Clothing and the Human Body: In the physics simulation engine, based on the pattern topology and fabric physical parameters, a bidirectional collision detection mechanism is constructed between clothing vertices and the virtual human body mesh surface. Each clothing vertex... With human face element The following constraints are defined between them: Penetration depth tolerance : The maximum allowable overlap distance in the normal direction, ranging from 1mm to 3mm.

[0042] sliding friction threshold The minimum coefficient of friction required for clothing to generate tangential motion on the contact surface; When the actual distance between the vertex and the surface element satisfy Then, the simulation engine automatically activates the rebound force and friction damping during the penetration processing stage.

[0043] S33, driving the virtual human to perform target actions: Based on a preset sequence of human actions (such as walking, turning, stretching, etc.), the virtual human is guided to produce dynamic posture changes by driving the rotation of skeletal nodes. The human skeleton transformation is achieved through the following rotation-translation structure: ; in For the first The time evolution transformation matrix of each joint This represents the antisymmetric matrix corresponding to the rotation vector. This is the initial transformation. After the driver executes, the simulation engine records the temporal displacement trajectory of each vertex on the virtual human body surface. , used for subsequent bonding calculations.

[0044] S34, Dynamic Fitting Displacement Field Solution: During frame-by-frame simulation, the fitting response displacement field of the clothing vertex is solved in each frame by combining the deformation trajectory of the virtual human body surface with the clothing-human body collision constraint relationship.

[0045] For each clothing vertex Its relative to the current contact surface Dynamic fitting displacement Decomposed into: ; Normal components Used to eliminate penetration. The normal vector of the contact surface; Tangential component : Reflects frictional slip, The relative sliding direction; coefficient The dynamic fit is determined by the fabric stiffness and coefficient of friction, and is automatically matched through the material library configuration in the simulation engine. For example, if a vertex initially contacts a thigh surface element, a 2mm penetration is detected, and the fabric is elastic denim, the dynamic fit displacement will bounce in the normal direction with a certain amount of sliding to simulate the dynamic effect of real tight-fitting clothing. S35, Garment Vertex-Driven Deformation: The calculated fitting displacement field is applied to the vertex set of the structured 3D garment model, such that: ; After frame-by-frame iterative updates, the structured 3D clothing model synchronously generates visible deformation and dynamic wrinkling effects in accordance with human movements. The simulation engine also records physical quantities such as fabric elongation and stress distribution, providing a basis for subsequent evaluation and optimization.

[0046] Ultimately, the goal is to achieve a natural response of clothing in dynamic environments, presenting a realistic fit, motion simulation, and physical plausibility in the way clothing is worn.

[0047] S4 includes: S41, Scene Semantic Tag Parsing: Receives user-input scene semantic tags, such as "summer street photography," "business meeting," and "outdoor rainy day," and performs semantic decomposition to obtain the following three dimensions: Ambient lighting requirements, including indoor / outdoor and natural / artificial light; Accessory functionality requirements, such as needing to match with handbags, umbrellas, and earrings; Background space requirements, such as city streets, conference halls, forest trails, etc.

[0048] Embedded, a semantic embedding model based on BERT The labels are vectorized and then input into the classification network to achieve the decomposition of semantic attribute dimensions: ; Output three semantic vectors as query conditions for subsequent material matching.

[0049] S42, Clothing Style and Physical Parameter Extraction: Extract the following two types of key features from the structured 3D clothing model; Clothing style feature vector: including color distribution histogram, obtained by HSV channel quantization statistics; and outline geometric style encoding, generated by ShapeNet principal component compression, used to capture local structural differences such as sleeve type, collar type, and hem; Physical parameter constraint set: including fabric light transmittance threshold And the load-bearing limit of accessories ,in; Fabric light transmittance threshold Used to limit excessive backlighting and the weight-bearing limit of accessories. Limit the maximum weight of accessories that can be attached.

[0050] These features are used to constrain the range of material matching to ensure the compatibility and physical plausibility of the generated scene; S43, Hierarchical Matching Search: Perform hierarchical matching search in the preset material library according to the three-dimensional semantic requirements and clothing characteristics: Level 1, Lighting Parameter Matching: Based on ambient lighting requirements Match the following set of lighting parameters: color_temp,attenuation_curve,shadow_sharpness ; Matching is achieved using the semantic vector nearest neighbor method, that is: ; The second level is accessory model matching: First, a set of accessory models that meet the functional requirements is selected. Then, a safety screening is performed based on the set of physical parameter constraints, with the following limitations: ; in, This represents the gravity value corresponding to the static load mass of the accessory model.

[0051] Level 3, Background Model Matching: This involves calculating the clothing color histogram. Histogram of background material colors Similarity to Bhattacharyya: ; like To meet the visual compatibility criteria, the clothing geometric style code and the background style description vector are combined and weighted to obtain the best matching background model. S44, Matching Element Binding and Spatial Configuration: Combine the above matching results into a complete matching element combination, including a lighting parameter set, an accessory model set, and a background model; For each accessory model, based on the predefined attachment point coordinates in the structured 3D clothing model Local coordinate system of accessories Perform the following spatial transformation binding: ; This transformation accurately embeds the accessory model into the attachment point of the clothing model, ensuring synchronous response during movement.

[0052] S45, 3D Display Scene Integration and Physical Interaction Generation: Load the following components in the physical simulation engine: Parametric virtual human body (including skeleton-driven and soft tissue deformation models); Structured 3D clothing model (including fabric physical properties and pattern topology); Combine elements (accessories, lighting, background).

[0053] The engine executes the following process in each frame: Update skeletal posture to drive human movement; Calculate the physical response between clothing and the human body; Real-time rendering of optical effects such as shadows, light transmission, and reflections in illuminated scenes; Accessories and clothing are rigidly bound or elastically suspended, responding to inertia and collision feedback; The background model can be fixed or dynamically adjusted to allow for interactive scene roaming and camera switching.

[0054] The final output features dynamic physical deformation, light and shadow changes, accessory interaction, and immersive scene, enhancing the immersive expression of clothing design solutions and user engagement.

[0055] like Figure 2 As shown, an intelligent display system for clothing design, used to implement the aforementioned intelligent display method for clothing design, includes the following modules: Sketch processing module: Used to receive clothing design sketches input by users, extract contour features through convolutional neural networks, and generate a structured 3D clothing model including pattern topology and fabric physical parameters; Virtual human body modeling module: used to acquire user body shape scan data and construct a parametric virtual human body based on biomechanical characteristics. The parametric virtual human body has a dynamic correlation model between the range of motion of skeletal joints and the deformation coefficient of soft tissue. Clothing Fitting Simulation Module: This module maps structured 3D clothing models onto parametric virtual human body surfaces and establishes collision constraint relationships in the physics simulation engine to achieve dynamic fit and physical deformation of clothing as the body moves. Scene generation module: used to parse the scene semantic tags input by the user, and based on clothing style features and physical parameter constraints, match lighting parameters, accessory models and background models in the material library to generate an interactive 3D display scene.

[0056] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0057] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An intelligent display method for clothing design, characterized in that, Includes the following steps: S1: Receive clothing design sketches, extract the outline features of the clothing design sketches through convolutional neural networks, and generate a structured 3D clothing model. The structured 3D clothing model includes pattern topology and fabric physical parameters. S2: Acquire user body shape scan data and generate a parameterized virtual human body based on biomechanical features, where the range of motion of skeletal joints is related to the soft tissue deformation coefficient; S3: Map the structured 3D clothing model onto the parameterized virtual human body surface, and calculate the dynamic fitting displacement field between the clothing vertices and the human body mesh through a real-time physics simulation engine to drive the clothing to deform with the human body's movements. S4: Parse the scene semantic tags input by the user, match matching elements from the material library that are compatible with the current clothing style and physical parameters, and generate an interactive 3D display scene including lighting, accessories and background.

2. The intelligent display method for clothing design according to claim 1, characterized in that, S1 includes: S11: Receive the clothing design sketch input by the user, and perform grayscale normalization and noise filtering preprocessing on the clothing design sketch; S12: Input the pre-processed clothing design sketch into a pre-trained convolutional neural network, and extract the multi-scale fusion outline features of the clothing design sketch through the encoder-decoder structure of the convolutional neural network. S13: Based on the outline features of the clothing design sketch, a structured 3D clothing model initial framework including NURBS surface control points is generated using a parametric surface reconstruction algorithm.

3. The intelligent display method for clothing design according to claim 2, characterized in that, S1 further includes: S14: Perform pattern segmentation on the initial frame of the structured 3D clothing model, and establish the pattern topology relationship between each pattern through a geometric constraint solver. The pattern topology relationship includes sewing connection relationship and overlapping constraint relationship. S15: Match the physical parameters of the fabric associated with the outline features of the garment design sketch from the preset fabric database. The physical parameters of the fabric include the bending stiffness coefficient, tensile elastic modulus, and coefficient of friction. S16: Bind the topological relationship of the pattern pieces and the physical parameters of the fabric to the initial framework of the structured 3D clothing model to generate a structured 3D clothing model with physical attributes.

4. The intelligent display method for clothing design according to claim 3, characterized in that, S2 includes: S21: Obtain user body shape scanning data through a 3D scanning device, and perform point cloud registration and hole repair processing on the user body shape scanning data; S22: Extract biomechanical features from the processed user body shape scan data, including bone point location markers, joint rotation axis vectors, and body surface muscle distribution heatmaps. S23: Based on the biomechanical characteristics, drive the deformation of the parameterized human body template to generate a parameterized virtual human body base mesh with anatomical hierarchical structure, and calculate the soft tissue deformation coefficients at the main joints.

5. The intelligent display method for clothing design according to claim 4, characterized in that, S2 further includes: S24: Establish a dynamic correlation model between the range of motion of skeletal joints and the deformation coefficient of soft tissue, wherein the range of motion of skeletal joints is defined by a rotation angle threshold matrix, and the soft tissue deformation coefficient is updated in real time by interpolation with the joint rotation angle. S25: Encode the dynamic association model into the vertex weight graph of the parameterized virtual human body base mesh to generate a parameterized virtual human body that can drive soft tissue deformation.

6. The intelligent display method for clothing design according to claim 5, characterized in that, S3 includes: S31: Spatially align the structured 3D clothing model with the parameterized virtual human body, and use a rigid body transformation matrix to make the initial position of the structured 3D clothing model fit the surface of the parameterized virtual human body. S32: Based on the topological relationship of the pattern and the physical parameters of the fabric, establish the collision constraint relationship between the vertices of the clothing and the surface of the human body mesh in the physical simulation engine. The collision constraint relationship includes the sliding friction threshold and the penetration depth tolerance. S33: Drive the parameterized virtual human body to perform the target action, and calculate the deformation of the parameterized virtual human body surface caused by the movement of the skeletal joints in real time through the physical simulation engine.

7. The intelligent display method for clothing design according to claim 6, characterized in that, S3 further includes: S34: Based on the surface deformation and collision constraint relationship, solve the dynamic fitting displacement field of the clothing vertex relative to the parameterized virtual human body surface frame by frame in the physical simulation engine. The dynamic fitting displacement field includes the vertex normal displacement component and the tangential sliding component. S35: Apply the dynamic fitting displacement field to the vertices of the structured 3D clothing model to drive the structured 3D clothing model to undergo physical deformation as the parametric virtual human body moves.

8. The intelligent display method for clothing design according to claim 7, characterized in that, S4 includes: S41: Parse the scene semantic tags input by the user, and decompose the scene semantic tags into three dimensions: ambient lighting requirements, accessory function requirements, and background space requirements; S42: Extract clothing style feature vectors and physical parameter constraint sets based on the structured 3D clothing model. The clothing style feature vectors include color distribution histograms and outline geometric style codes. The physical parameter constraint sets include fabric light transmittance thresholds and accessory load-bearing limits. S43: Perform multi-level matching search in the preset material library, including: Level 1: Match lighting parameters according to the ambient lighting requirements, including light source color temperature, intensity attenuation curve, and shadow sharpness; Level 2: Select accessory models based on the accessory functional requirements and physical parameter constraint set, ensuring that the load-bearing value of the accessory model's attachment point is less than the accessory load-bearing limit in the clothing physical parameter constraint set; Level 3: Retrieve the background model based on the background space requirements and clothing style feature vectors, requiring the color distribution of the background model and the histogram of clothing color distribution to meet the visual compatibility criteria.

9. The intelligent display method for clothing design according to claim 8, characterized in that, S4 further includes: S44: Combine the matched lighting parameters, accessory models, and background models into a matching element combination, and bind the accessory models to the corresponding attachment points of the structured 3D clothing model through spatial coordinate transformation; S45: Integrate the structured 3D clothing model, parametric virtual human body and matching element combination into the physical simulation engine to generate an interactive 3D display scene with dynamic lighting effects and physical interaction.

10. An intelligent display system for clothing design, used to implement the intelligent display method for clothing design as described in any one of claims 1-9, characterized in that, Includes the following modules: Sketch processing module: Used to receive clothing design sketches input by users, extract contour features through convolutional neural networks, and generate a structured 3D clothing model including pattern topology and fabric physical parameters; Virtual human body modeling module: used to acquire user body shape scanning data and construct a parameterized virtual human body based on biomechanical characteristics. The parameterized virtual human body has a dynamic correlation model between the range of motion of skeletal joints and the deformation coefficient of soft tissue. Clothing Fitting Simulation Module: This module maps structured 3D clothing models onto parametric virtual human body surfaces and establishes collision constraint relationships in the physics simulation engine to achieve dynamic fit and physical deformation of clothing as the body moves. Scene generation module: used to parse the scene semantic tags input by the user, and based on clothing style features and physical parameter constraints, match lighting parameters, accessory models and background models in the material library to generate an interactive 3D display scene.

Images