A 2D garment crease dynamic fitting method and system based on topological interpolation

Abstract

The application discloses a 2D garment crease dynamic fitting method and system based on topological interpolation, belongs to the field of computer graphics image processing, and comprises the following steps: obtaining garment mesh and skeleton motion data; performing dynamic physical parameterization on the mesh; calculating a dynamic energy distribution field through a topological interpolation algorithm; activating and synthesizing macro-creases from a wrinkle form dictionary based on the energy field; generating secondary detailed creases according to the spatial gradient of the energy field; and performing time sequence consistency fusion and rendering processing. The application adopts a nonlinear synthesis technology combining energy conduction simulation based on topological connection relationship and manifold learning dictionary matching, can simulate the crease evolution process of different material fabrics under dynamic motion, and significantly improves the physical reality, visual detail richness and motion stability of 2D garment creases.

Description

Technical Field

[0001] This invention relates to the field of computer graphics image processing, and in particular to a method and system for dynamic fitting of 2D clothing creases based on topological interpolation. Background Technology

[0002] Against the backdrop of the rapid development of the digital culture industry and virtual reality technology, the dynamic effects of virtual clothing have become a research hotspot in the field of computer graphics. In applications such as 3D animation, video games, and virtual try-on, it is necessary to accurately simulate the dynamic changes of clothing creases during character movement in order to enhance visual realism.

[0003] In related technologies, Chinese invention patent with announcement number CN116071500A discloses a clothing design method and system based on 3D modeling. The method includes inputting target object information and fabric material attribute information into a fabric difference analysis model, outputting 3D virtual adjustment parameters for 3D modeling, generating 2D clothing disassembly information using a dimension conversion module and inputting it into a topology carving module, and finally using a rendering module to obtain the designed clothing.

[0004] Regarding the aforementioned technologies, the inventors believe that although 3D modeling technology can be used to realize clothing design, when simulating dynamic folds in clothing, traditional 3D physical simulation involves extremely complex mesh deformation calculations and collision detection, resulting in a huge computational load and high data processing complexity. This high computational cost makes it difficult for dynamic fold generation to meet the real-time requirements of interactive applications, and it is prone to visual jumps and artifacts under high-speed motion, which is not conducive to the stable output of high-fidelity dynamic effects. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a 2D clothing crease dynamic fitting method and system based on topological interpolation. The method employs a technique that calculates the dynamic energy distribution field on a two-dimensional clothing pattern mesh through topological interpolation and combines it with a pre-constructed crease morphology dictionary for nonlinear synthesis. This approach can generate dynamic clothing crease effects that combine physical realism with rich visual details while ensuring high operating efficiency.

[0006] To achieve the above objectives, this application adopts the following technical solution:

[0007] Firstly, a method for dynamically fitting 2D clothing creases based on topological interpolation is provided, comprising: acquiring a two-dimensional polygonal mesh of the garment to be processed and motion data of the character skeleton driving the garment; dynamically physical parameterizing the two-dimensional polygonal mesh, assigning each cell in the mesh a set of dynamic physical attributes including basic deformation stiffness and energy conduction rate factor; calculating a dynamic energy distribution field covering the entire garment area using a topological interpolation algorithm based on the character skeleton motion data, the dynamic physical attribute set, and the topological connectivity of the mesh; performing nonlinear synthesis of crease morphology based on the dynamic energy distribution field, wherein, by comparing the energy intensity of the dynamic energy distribution field with a crease generation threshold dynamically determined according to material properties, crease geometric patterns are activated and synthesized from a crease morphology dictionary pre-constructed through manifold learning to generate macroscopic crease geometric data; generating secondary detailed crease data based on the spatial gradient of the dynamic energy distribution field and the macroscopic crease geometric data; performing temporal consistency fusion and rendering processing on the macroscopic crease geometric data and the secondary detailed crease data to output a dynamic clothing crease image.

[0008] Based on the above technical solution, in the 2D clothing crease dynamic fitting method based on topological interpolation provided in this application, the dynamic energy distribution field is calculated by topological interpolation on the two-dimensional clothing pattern mesh, and nonlinear synthesis is performed in combination with a pre-constructed crease morphology dictionary. This method can generate dynamic clothing crease effects with both physical realism and rich visual details while ensuring high operating efficiency.

[0009] In conjunction with the first aspect above, in one possible implementation, the dynamic physical parameterization of the two-dimensional pattern polygon mesh includes: configuring an initial physical attribute set containing standardized parameter values ​​for each cell of the two-dimensional pattern polygon mesh; obtaining a fabric type identifier for specifying the type of garment fabric to be processed; retrieving a corresponding reference physical parameter range from a database storing several real fabric physical properties based on the fabric type identifier; and mapping and adjusting the parameters in the initial physical attribute set according to the reference physical parameter range to generate a dynamic physical attribute set.

[0010] In conjunction with the first aspect above, in one possible implementation, the calculation of the dynamic energy distribution field covering the entire clothing area using a topological interpolation algorithm, based on the character skeleton motion data, dynamic physical attribute set, and the topological connection relationship of the mesh, includes: calculating the virtual force applied to each vertex of the two-dimensional polygonal mesh based on the joint rotation information in the character skeleton motion data; constructing a topological relationship model to describe the energy conduction relationship between mesh units based on the energy conduction rate factor in the dynamic physical attribute set; and simulating the energy conduction and superposition process on the two-dimensional polygonal mesh based on the virtual force and the topological relationship model, thereby calculating the dynamic energy distribution field composed of energy intensity and energy direction vectors.

[0011] In conjunction with the first aspect above, in one possible implementation, the nonlinear synthesis of fold morphology based on the dynamic energy distribution field includes: performing nonlinear topological interpolation on the dynamic energy distribution field over a time series based on the topological relationship model to obtain an energy field evolution sequence containing energy intensity evolution and direction vector evolution; identifying a set of mesh vertices in the energy field evolution sequence whose energy intensity exceeds the fold generation threshold to form fold trigger points; matching several candidate fold primitives from the fold morphology dictionary pre-constructed through manifold learning based on the energy intensity value and energy direction vector at the fold trigger points; performing geometric transformation and intensity modulation on the matched candidate fold primitives and superimposing them onto the two-dimensional polygonal mesh to generate macroscopic fold geometric data.

[0012] In conjunction with the first aspect above, in one possible implementation, the wrinkle generation threshold dynamically determined based on material properties includes: obtaining the basic deformation stiffness value of each mesh unit from the dynamic physical property set; mapping the basic deformation stiffness value to an energy threshold through a preset functional relationship; and using the energy threshold as the wrinkle generation threshold, wherein a high basic deformation stiffness value is mapped to a high wrinkle generation threshold, and a low basic deformation stiffness value is mapped to a low wrinkle generation threshold.

[0013] In conjunction with the first aspect above, in one possible implementation, the fold morphology dictionary includes: acquiring images of a sample garment under different motion states to construct a two-dimensional fold sample set; performing manifold learning dimensionality reduction processing on the two-dimensional fold sample set to extract a low-dimensional fold feature manifold; performing basis decomposition on the fold feature manifold to identify several basis vectors representing typical fold patterns; associating the several basis vectors with their corresponding energy aggregation patterns and storing them as fold primitives in a library to construct a fold morphology dictionary.

[0014] In conjunction with the first aspect above, in one possible implementation, generating secondary detail wrinkle data based on the macroscopic wrinkle geometry data according to the spatial gradient of the dynamic energy distribution field includes: calculating the gradient vector and gradient magnitude of the dynamic energy distribution field at each vertex of the two-dimensional polygonal mesh; selecting vertices whose gradient magnitude exceeds a preset detail generation threshold used to control detail density, forming detail generation regions; generating high-frequency perturbation signals with frequencies and amplitudes related to the gradient magnitude based on the gradient vector direction at the detail generation regions; and converting the high-frequency perturbation signals into microscopic geometric displacements to generate secondary detail wrinkle data.

[0015] In conjunction with the first aspect above, in one possible implementation, the temporal consistency fusion and rendering processing of the macroscopic wrinkle geometry data and the secondary detail wrinkle data includes: fusing the macroscopic wrinkle geometry data and the secondary detail wrinkle data of the current frame to obtain the total wrinkle geometry data of the current frame; obtaining the total wrinkle geometry data generated after processing the previous frame; performing smoothing filtering processing on the total wrinkle geometry data of the current frame and the total wrinkle geometry data after processing the previous frame using a temporal convolution kernel to obtain smoothed total wrinkle geometry data; converting the smoothed total wrinkle geometry data into normal map information, and combining it with the basic clothing texture for rendering to output a dynamic clothing wrinkle image.

[0016] In conjunction with the first aspect above, in one possible implementation, the method further includes: real-time monitoring of the character's skeletal motion data, calculating a global motion intensity index that quantifies the severity of its changes; comparing the global motion intensity index with a stability threshold used to distinguish motion states; and adaptively enhancing the smoothing intensity of the temporal convolution kernel when the global motion intensity index exceeds the stability threshold to maintain the visual stability of the wrinkles.

[0017] Secondly, a 2D clothing crease dynamic fitting system based on topological interpolation is provided, comprising: a data acquisition module for acquiring a two-dimensional polygonal mesh of the clothing to be processed and motion data of the character skeleton driving the clothing; a physical parameterization module for dynamically physical parameterizing the two-dimensional polygonal mesh, assigning each cell in the mesh a set of dynamic physical attributes including basic deformation stiffness and energy conduction rate factor; an energy field calculation module for calculating the dynamic energy distribution field covering the entire clothing area using a topological interpolation algorithm based on the character skeleton motion data, the dynamic physical attribute set, and the topological connectivity of the mesh; and a macroscopic wrinkle synthesis module. This module is used for nonlinear synthesis of fold patterns based on the dynamic energy distribution field. It compares the energy intensity of the dynamic energy distribution field with a fold generation threshold dynamically determined according to material properties, activates and synthesizes fold geometric patterns from a pre-constructed fold pattern dictionary using manifold learning, and generates macroscopic fold geometric data. A detailed fold generation module generates secondary detailed fold data based on the spatial gradient of the dynamic energy distribution field and the macroscopic fold geometric data. A fusion rendering module performs temporal consistency fusion and rendering processing on the macroscopic fold geometric data and the secondary detailed fold data, outputting a dynamic clothing fold image.

[0018] Compared with the prior art, the present invention has the following advantages:

[0019] This invention simplifies the complex three-dimensional physical simulation process of clothing by transforming it into dynamic energy field calculation on a two-dimensional clothing pattern mesh. This reduces the computational load and data processing complexity of the system, making the generation of dynamic folds more efficient and meeting the real-time requirements of interactive applications.

[0020] This invention combines a physical law-based energy conduction model with data-driven fold morphology synthesis. By obtaining physical parameters from a real fabric database and using a fold morphology dictionary constructed using manifold learning, it ensures that the generated folds not only follow physical logic but also possess rich nonlinear features and appearances derived from real samples, thereby enhancing visual realism.

[0021] This invention employs a multi-layered detail generation strategy, dividing wrinkles into macroscopic structures driven by energy intensity and secondary details driven by energy gradients. Temporal consistency fusion ensures smooth and natural dynamic changes. This method can generate the main wrinkle outlines while supplementing rich high-frequency microscopic details, effectively suppressing visual jumps and artifacts during animation, thus achieving stable and high-fidelity dynamic rendering effects.

[0022] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A structural architecture diagram of a 2D clothing crease dynamic fitting system based on topological interpolation provided in this application embodiment;

[0025] Figure 2 A flowchart illustrating a 2D clothing crease dynamic fitting method based on topological interpolation provided in this application embodiment;

[0026] Figure 3 This is a dynamic energy distribution field diagram provided in the embodiments of this application.

[0027] Figure 4 This is a smoothing coefficient adaptive adjustment mapping curve provided in the embodiments of this application. Detailed Implementation

[0028] In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. The "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0029] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0030] The 2D clothing crease dynamic fitting method based on topological interpolation provided in this application embodiment can be applied to, for example... Figure 1 In the 2D clothing crease dynamic fitting system 100 based on topological interpolation shown, as... Figure 1 As shown, the system includes:

[0031] The data acquisition module is used to acquire the two-dimensional polygonal mesh of the garment to be processed and the skeletal motion data of the character driving the garment;

[0032] The physical parameterization module is used to dynamically physical parameterize the two-dimensional polygonal mesh, assigning each cell in the mesh a set of dynamic physical properties including basic deformation stiffness and energy conduction rate factor.

[0033] The energy field calculation module is used to calculate the dynamic energy distribution field covering the entire clothing area based on the character's skeletal motion data, dynamic physical attribute set, and the topological connection relationship of the mesh, using a topological interpolation algorithm.

[0034] The macroscopic fold synthesis module is used to perform nonlinear synthesis of fold morphology based on the dynamic energy distribution field. The module compares the energy intensity of the dynamic energy distribution field with a fold generation threshold dynamically determined according to the material properties, activates and synthesizes fold geometric patterns from a fold morphology dictionary pre-constructed through manifold learning, and generates macroscopic fold geometric data.

[0035] The detail wrinkle generation module is used to generate secondary detail wrinkle data based on the macroscopic wrinkle geometry data according to the spatial gradient of the dynamic energy distribution field.

[0036] The fusion rendering module is used to perform temporal consistency fusion and rendering processing on the macroscopic fold geometric data and the secondary detailed fold data, and output dynamic clothing fold images.

[0037] like Figure 2 As shown, this application provides a 2D clothing crease dynamic fitting method based on topological interpolation, including:

[0038] Obtain the 2D polygonal mesh of the garment to be processed and the skeletal motion data of the character driving the garment;

[0039] The two-dimensional polygonal mesh is dynamically physically parameterized, and each cell in the mesh is assigned a set of dynamic physical properties including basic deformation stiffness and energy conduction rate factor.

[0040] Based on the character's skeletal motion data, dynamic physical attribute set, and the topological connection relationship of the mesh, a dynamic energy distribution field covering the entire clothing area is calculated using a topological interpolation algorithm.

[0041] Nonlinear synthesis of fold morphology is performed based on the dynamic energy distribution field, wherein the energy intensity of the dynamic energy distribution field is compared with a fold generation threshold dynamically determined according to material properties, and fold geometric patterns are activated and synthesized from a fold morphology dictionary pre-constructed through manifold learning to generate macroscopic fold geometric data.

[0042] Based on the spatial gradient of the dynamic energy distribution field, secondary detailed fold data are generated on the basis of the macroscopic fold geometry data;

[0043] The macroscopic fold geometric data and the secondary detailed fold data are subjected to temporal consistency fusion and rendering processing to output a dynamic clothing fold image.

[0044] It should be noted that a data-driven and physically simplified model is constructed, using a dynamic energy distribution field as a medium to connect the character's skeletal motion with the generation of 2D clothing folds. First, the 2D geometric pattern of the clothing is transformed into a dynamic mesh with physical properties such as basic deformation stiffness and energy conduction rate. Next, the character's skeletal motion data is parsed into virtual forces applied to this mesh, and a topological interpolation algorithm is used to simulate the conduction and distribution of this energy across the clothing pattern according to the set physical properties, forming a dynamic energy distribution field covering the entire clothing area. Fold generation is decomposed into two levels: macroscopic and detailed. Macroscopic folds are non-linearly activated and synthesized from a pre-constructed fold morphology dictionary built through manifold learning by comparing the intensity of the energy field with a dynamic threshold; secondary detailed folds are generated based on the spatial gradient of the energy field. Finally, the fold data from both levels are fused and temporally smoothed before the final rendering output.

[0045] In one possible implementation of the embodiments of this application, combined with Figure 2 The dynamic physical parameterization of the two-dimensional polygonal mesh includes:

[0046] Configure each cell of the two-dimensional polygonal mesh with an initial set of physical properties containing standardized parameter values;

[0047] Obtain the fabric type identifier used to specify the type of garment fabric to be processed;

[0048] Based on the fabric type identifier, retrieve the corresponding reference physical parameter range from a database that stores the physical properties of several real fabrics.

[0049] Based on the reference physical parameter range, the parameters in the initial physical attribute set are mapped and adjusted to generate a dynamic physical attribute set.

[0050] In some implementations, the two-dimensional polygonal mesh is dynamically physically parameterized, transforming the abstract geometric mesh into a physical model capable of responding to external drives and simulating the behavior of real fabrics, laying the foundation for subsequent energy field calculations. This is accomplished by configuring an initial physical attribute set containing standardized parameter values ​​for each cell of the two-dimensional polygonal mesh. All polygonal cells in the mesh are traversed, and an initial physical attribute set is instantiated in the data structure of each cell. This attribute set contains at least two key parameters: basic deformation stiffness and energy conduction rate factor. The initial values ​​are set to standardized floating-point numbers between 0 and 1, such as 0.5, thus establishing a material-neutral, unified physical benchmark. Based on the user-specified fabric type, precise parameters conforming to real-world physical laws are retrieved from a pre-built physical property library as the basis for adjustment. A fabric type identifier is obtained to specify the fabric type to be processed; this identifier can be a string, such as "denim" or "silk". This fabric type identifier is used as a query index to search a database storing the physical properties of several real fabrics. Each entry in this database corresponds to a specific fabric and stores a range of reference physical parameters for its physical properties. For example, the basic deformation stiffness range for denim might be... The energy conduction rate factor ranges from 100 to 100. The reference physical parameter range matching the fabric type identifier is retrieved to provide data support for subsequent parameter mapping. The standardized initial physical parameters are mapped to the physical parameter range of the real fabric, generating the final dynamic physical attribute set for dynamic simulation. Based on the obtained reference physical parameter range, the parameters in the initial physical attribute set of each mesh cell are mapped and adjusted. This mapping process is implemented through the following function, generating the final dynamic physical attribute set. The calculation formula is: ;in, This represents the value of a specific parameter in the final generated dynamic physical property set, such as the adjusted basic deformation stiffness. It is the standardized value of the parameter in the initial physical property set. and These are the lower and upper limits of the reference physical parameter range retrieved from the fabric database, respectively. This function linearly scales the standardized initial values ​​to the range of physical properties of the actual fabric, thereby ensuring that the differences in physical behavior between different fabrics are accurately reflected. The final output set of dynamic physical properties will be directly applied to subsequent energy distribution field calculations.

[0051] For example, assuming a garment with fabric labeled "silk", first assign an initial set of physical properties to a cell of the two-dimensional pattern mesh, assuming a standardized initial value for its basic deformation stiffness. The value is 0.5. Subsequently, the reference physical parameter range corresponding to silk material, i.e., the lower limit of the basic deformation stiffness, was retrieved from the database. The upper limit is 0.1. A value of 0.3 reflects the soft and easily deformable characteristics of silk. Finally, through the mapping function... Calculations are performed, and the final basic deformation stiffness of the element is obtained by substituting the numerical values. Through this mapping calculation, a neutral geometric mesh was successfully transformed into a dynamic model with specific physical properties of silk, achieving a precise alignment from abstract numerical values ​​to the physical behavior of the fabric.

[0052] In one possible implementation, combining Figure 2 Based on the character's skeletal motion data, dynamic physical attribute set, and mesh topological connectivity, a dynamic energy distribution field covering the entire clothing area is calculated using a topological interpolation algorithm, including:

[0053] Based on the joint rotation information in the character skeleton motion data, calculate the virtual force applied to each vertex of the two-dimensional polygonal mesh.

[0054] Based on the energy conduction rate factor in the dynamic physical attribute set, a topological relationship model is constructed to describe the energy conduction relationship between grid cells;

[0055] Based on the virtual force and the topological relationship model, the transmission and superposition process of energy on the two-dimensional polygonal mesh is simulated, and the dynamic energy distribution field composed of energy intensity and energy direction vector is calculated.

[0056] In some implementations, a dynamic energy distribution field covering the entire garment area is calculated using a topological interpolation algorithm based on character skeletal motion data, a dynamic physical attribute set, and the topological connectivity of the mesh. This transforms discrete skeletal motion into a continuous energy data field distributed across the garment surface, providing a physical driving force for subsequent wrinkle synthesis. Virtual forces applied to each vertex of the 2D polygonal mesh are calculated based on joint rotation information from the character skeletal motion data. By analyzing the character skeletal motion data, the angle changes and angular velocities of key joints, such as elbows or knees, between the current and previous frames are obtained. For each mesh vertex affected by joint motion, the kinetic energy of the joint is converted into a virtual force vector based on its skinning weight. The magnitude of this vector is positively correlated with the intensity of the joint motion, and its direction is consistent with the tangential motion direction caused by joint rotation. This virtual force represents the initial injection point and intensity of kinetic energy. A mathematical model describing how energy is conducted within the garment mesh is established, quantifying the physical properties of the fabric into rules for energy flow. Based on the energy conduction rate factor in the generated dynamic physical attribute set, a topological relationship model describing the energy conduction relationship between mesh cells is constructed. This topological model is essentially a weighted graph, where nodes are cells of a two-dimensional polygonal mesh, and edges represent the adjacency relationships between cells. The weight of each edge is determined by the energy conduction rate factors of the two cells it connects to; for example, the average or minimum of the two factor values ​​can be used. A higher weight means that energy can be transferred between the two cells more losslessly and quickly, which aligns with the physical characteristics of high-stiffness or high-density fabrics. Through simulation calculations, the final distribution state of energy after propagation across the entire garment mesh according to the topological model and reaching quasi-static equilibrium under the initial virtual force is determined. Based on the calculated virtual force and the constructed topological model, an iterative topological interpolation algorithm simulates the energy conduction and superposition process on the two-dimensional polygonal mesh. In each iteration step, the energy of each mesh cell diffuses to its neighboring cells according to the weights defined by the topological model, while also receiving energy from neighboring cells. The energy update of this process can be expressed as: ;in, It is a grid cell The energy vector at the next moment. and They are units and its neighboring units The energy vector at the current moment. It is a global energy decay coefficient ranging from 0.05 to 0.2, used to simulate energy dissipation. It is the transmission scale factor. It is a unit defined in the topological relationship model. and The transmission weights between them. After several iterations, typically 5 to 20, the energy distribution tends to stabilize. The final output dynamic energy distribution field is a vector field, where each grid cell contains a complete energy state description composed of the energy intensity and energy direction vectors, i.e., the magnitude of the energy vector and the unit vector of the energy vector. Figure 3 As shown, the topological conduction of energy on a two-dimensional grid is illustrated by contour lines. The numerical values ​​represent energy intensity and reflect the attenuation logic of energy diffusion from the virtual force injection point to the entire domain.

[0057] For example, when a character's arm bends to drive the sleeve, the initial virtual force vector applied to the apex of the sleeve cuff area is first calculated based on the angular velocity generated by the rotation of the elbow joint. Let its energy vector be... The intensity is 10.0. Subsequently, the transmission weights between units are determined based on the topological relationship model. A value of 0.8 indicates a relatively dense fabric, and a global energy attenuation coefficient is set. The transmission scaling factor is 0.1. The value is 0.2. In the first round of topological interpolation iteration, this cell receives conducted energy from neighboring cells, assuming the neighbor energy... The value is 5.0, so we substitute it into the formula. Calculations were performed to obtain After multiple rounds of such iterative calculations, energy smoothly diffuses from the elbow stress point along the mesh topology to the entire sleeve, ultimately forming a dynamic energy distribution field containing precise intensity and direction information, providing a quantitative physical driving force for the subsequent activation of wrinkle primitives.

[0058] In one possible implementation, combining Figure 2 The nonlinear synthesis of fold morphology based on the dynamic energy distribution field includes:

[0059] Based on the aforementioned topological relationship model, nonlinear topological interpolation is performed on the dynamic energy distribution field over time to obtain an energy field evolution sequence that includes energy intensity evolution and direction vector evolution.

[0060] Identify the set of mesh vertices in the energy field evolution sequence whose energy intensity exceeds the wrinkle generation threshold, and form wrinkle trigger points;

[0061] Based on the energy intensity value and energy direction vector at the wrinkle trigger point, several candidate wrinkle primitives are matched from the wrinkle morphology dictionary pre-constructed through manifold learning;

[0062] The matched candidate fold primitives are geometrically transformed and intensity modulated, and then superimposed onto the two-dimensional polygonal mesh to generate macroscopic fold geometric data.

[0063] In some implementations, nonlinear synthesis of fold morphology is performed based on a dynamic energy distribution field, transforming abstract energy field data into concrete, visually believable macroscopic fold geometry. Based on a topological relationship model, nonlinear topological interpolation is performed on the dynamic energy distribution field over time to smooth and stabilize drastic fluctuations in the energy field and extract the dominant trend of energy change. The most recent frames, such as 3 to 5 frames of dynamic energy distribution field, are cached, and the topological relationship model is used as a weighting guide to perform weighted averaging or more complex interpolation calculations on the energy data over these time series, generating a more temporally coherent and spatially smoother energy field evolution sequence. This sequence not only includes the energy state of the current frame but also the evolutionary trend of energy intensity and direction vector over a short period. Based on the degree of energy accumulation, the specific locations where folds should begin to form on the garment surface are precisely identified. Each grid vertex in the energy field evolution sequence is traversed, and its energy intensity is compared with a pre-determined fold generation threshold dynamically based on material properties. When the energy intensity of a vertex persists for a period of time or significantly exceeds this threshold in the current frame, the vertex is marked as a fold trigger point, forming a set of fold trigger points. This threshold is a key criterion for distinguishing whether fabric buckles and wrinkles due to stress. Appropriate wrinkle appearance morphologies are matched to the identified wrinkle trigger points. Based on the energy intensity value and energy direction vector at each wrinkle trigger point, a query and match are performed from a pre-constructed wrinkle morphology dictionary built through manifold learning. This dictionary stores a series of standardized wrinkle primitives, each representing a typical wrinkle pattern and associated with a specific energy accumulation pattern. Nearest neighbor search or similarity matching algorithms are used to find several candidate wrinkle primitives in the dictionary's feature space that best match the energy characteristics of the current trigger point. For example, a high-intensity, highly directional energy vector will match a linear, sharp crease primitive. The abstract wrinkle primitives are instantiated and applied to the garment mesh to complete the geometric construction of macroscopic wrinkles. The matched candidate wrinkle primitives undergo geometric transformation and intensity modulation. The geometric transformation includes rotating the primitives according to the energy direction vector to align their orientation with the energy flow direction, and scaling the primitive's length and influence range according to the energy intensity. Intensity modulation adjusts the height or depth of the primitives based on the energy intensity value; the stronger the energy, the more pronounced the wrinkles. The transformed and modulated wrinkle primitives are linearly superimposed onto the corresponding positions of the two-dimensional polygonal mesh through vertex displacement or normal perturbation, ultimately generating macroscopic wrinkle geometry data containing the main and large wrinkle shapes.

[0064] For example, the compression wrinkles generated at the elbow of a sleeve are first interpolated temporally from the energy fields of the most recent three frames in the cache to extract the dominant trend. Let the average energy intensity at a certain wrinkle trigger point be 15.0, and its dominant direction vector be... The wrinkle generation threshold, determined based on the fabric material, is 12.0. Since the current intensity significantly exceeds this threshold, the point is confirmed as activated. Subsequently, a standard "linear extrusion" wrinkle primitive is matched from the wrinkle morphology dictionary; this primitive has an initial height parameter of 1.0. Intensity modulation is performed based on the increment of the current energy intensity relative to the threshold, denoted as: ,in For the height of the primitive, This is the energy redundancy. With a scaling factor of 0.5, the actual instantiated height of the fold at that location is calculated as follows: Finally, based on the energy direction vector The calculated rotation angle is approximately By superimposing scaled and rotated geometric primitives onto a mesh, macroscopic creases that perfectly match the bending force and direction of the arm are precisely generated on a 2D plane.

[0065] In one possible implementation, combining Figure 2 The wrinkle generation threshold dynamically determined based on material properties includes:

[0066] Obtain the basic deformation stiffness value of each mesh element from the dynamic physical property set;

[0067] The basic deformation stiffness value is mapped to an energy threshold through a preset functional relationship;

[0068] The energy threshold is used as the wrinkle generation threshold, wherein a high foundation deformation stiffness value is mapped to a high wrinkle generation threshold, and a low foundation deformation stiffness value is mapped to a low wrinkle generation threshold.

[0069] In some implementations, a wrinkle generation threshold dynamically determined based on material properties is defined. This establishes differentiated, physically intuitive wrinkle formation standards for fabrics with different physical properties, ensuring that stiff fabrics require more energy to generate wrinkles, while soft fabrics wrinkle more easily. The basic deformation stiffness value of each grid cell is obtained from the dynamic physical property set. Before performing wrinkle morphology synthesis, for each grid cell, its dynamic physical property set, determined during the dynamic physical parameterization phase, is accessed, and the key parameter, the basic deformation stiffness value, is retrieved. This value is a direct indicator of the fabric's resistance to bending deformation. The physical deformation stiffness attribute is converted into energy units used in energy field calculations, establishing a direct mapping relationship from material properties to energy thresholds. The basic deformation stiffness value is mapped to the energy threshold through a preset functional relationship. This functional relationship ensures a positive correlation between stiffness and threshold. The linear mapping function is expressed as:

[0070] ;in, This represents the final calculated energy threshold. It is the basic deformation stiffness value of the mesh element obtained from the dynamic physical property set. It is a preset stiffness threshold mapping coefficient, a dimensionless, globally adjustable parameter, whose value range is usually between 1.5 and 3.0, used to control the sensitivity of stiffness to the threshold. This is the basic energy threshold, representing the minimum energy activation threshold required to form wrinkles without any stiffness influence. This function directly converts the physical properties of the material into values ​​that can be compared in the energy field. The calculated energy threshold is formally designated as the trigger condition for wrinkle synthesis, making it effective in subsequent calculations. The calculated energy threshold is used as the wrinkle generation threshold to determine whether a mesh element generates wrinkles. During subsequent nonlinear synthesis of wrinkle morphology, the energy intensity at the corresponding location in the dynamic energy distribution field is compared with this dynamically determined wrinkle generation threshold. Since high basic deformation stiffness values ​​are mapped to high wrinkle generation thresholds through the function, and low basic deformation stiffness values ​​are mapped to low wrinkle generation thresholds, high-stiffness fabrics such as denim require greater energy accumulation to trigger wrinkle generation, while low-stiffness fabrics such as silk can form wrinkles with less energy, thus achieving material-differentiated simulation of wrinkle behavior.

[0071] For example, comparing the triggering conditions of two fabrics, "heavy denim" and "thin silk," the basic deformation stiffness values ​​of each unit are first obtained from the physical property set. Set up denim fabric unit It is 0.8, while the silk fabric unit is for

[0072] 0.2. Then, the mapping function is applied. Threshold calculation is performed, where a stiffness threshold mapping coefficient is set. The base energy threshold is 2.5. The value is 5.0. For the denim unit, the wrinkle generation threshold is obtained by substituting the numerical value. For the silk unit, its threshold was calculated. This mapping process sets a higher trigger threshold of 7.0 for stiff denim fabric and a lower threshold of 5.5 for soft silk, allowing the same intensity of energy injection in subsequent calculations to trigger completely different wrinkle activation results, thus accurately simulating the unique mechanical response differences of materials at the 2D level.

[0073] In one possible implementation, combining Figure 2 The fold morphology dictionary includes:

[0074] Acquire images of the sample garment under different motion states, and construct a two-dimensional fold sample set;

[0075] Manifold learning dimensionality reduction is performed on the two-dimensional fold sample set to extract the low-dimensional fold feature manifold;

[0076] Basis decomposition is performed on the wrinkled feature manifold to identify several basis vectors representing typical wrinkling patterns;

[0077] The aforementioned basis vectors and their corresponding energy accumulation patterns are associated and stored in the library as fold primitives to construct a fold morphology dictionary.

[0078] In some implementations, a fold morphology dictionary is constructed, establishing a queryable database containing various basic fold patterns to provide geometric prototypes for subsequent nonlinear synthesis. A two-dimensional fold sample set is constructed by acquiring images of sample garments in different motion states. A series of pre-recorded or physically simulated high-precision garment animation sequences, covering various movements from slow walking to vigorous running and jumping, are loaded. For each frame in the sequence, the normal map or height map of the garment area is extracted and used as a two-dimensional geometric representation of the fold morphology at that moment, ultimately accumulating into a two-dimensional fold sample set containing hundreds to thousands of sample images. From the massive, high-dimensional fold sample data, the inherent, low-dimensional structural patterns are discovered, describing complex fold changes using a few key parameters. Manifold learning dimensionality reduction is performed on the two-dimensional fold sample set. Here, manifold learning is a nonlinear dimensionality reduction technique aimed at discovering low-dimensional manifold structures embedded in high-dimensional data space. Algorithms such as Isomap or LLE are used to map each high-dimensional wrinkle image sample as a high-dimensional vector to a point in a low-dimensional space, typically with dimensions between 10 and 30. The distribution of all sample points in this low-dimensional space constitutes a wrinkle feature manifold that can compactly describe all wrinkle variations. From the continuous feature manifold, the most representative typical wrinkle patterns that can serve as independent building blocks are identified and extracted. Basis decomposition is performed on the low-dimensional wrinkle feature manifold. This process typically employs techniques such as Principal Component Analysis (PCA) or Independent Component Analysis (ICA) to calculate the principal directions of the manifold space. These principal directions are several basis vectors, each geometrically corresponding to a basic, orthogonal, or statistically independent wrinkle pattern, such as lateral compression wrinkles or diagonal stretching wrinkles. Basis vectors whose contribution rate exceeds a certain threshold are retained, such as those with a cumulative contribution rate reaching a certain threshold. As the core constituent element of fold morphology, abstract geometric basis vectors are bound to the physical causes driving their generation, forming complete fold primitives that can be activated by the energy field. Several basis vectors are associated with their corresponding energy aggregation patterns. For each identified basis vector, the original two-dimensional fold sample set is traced back to find those samples with the strongest projection in the direction of that basis vector. Then, the original simulation data that generated these samples is analyzed to extract the common features of the corresponding dynamic energy distribution field, such as energy intensity, principal direction, and spatial distribution pattern, forming a characteristic energy aggregation pattern. The geometric shape represented by each basis vector and its associated energy aggregation pattern are packaged and stored as fold primitives in the library, thus completing the construction of the entire fold morphology dictionary.

[0079] For example, to extract radial folds under the armpits, a sample set is first constructed by obtaining 2000 height images of a pattern garment with different running and jumping movements. Then, the Isomap algorithm is used to reduce the dimensionality of each high-dimensional image to a 20-dimensional feature manifold space. Subsequently, basis decomposition, such as PCA, is performed on this manifold to extract the top 10 features with the highest cumulative contribution rate. The basis vectors, where the 3rd basis vector is... After geometric reconstruction, it displays a typical radial shape. Retrospective analysis revealed that samples with projection values ​​exceeding a threshold in the feature space corresponded to the character's arm-raising motion, such as the projection coefficient. At this time, the energy intensity of its energy distribution field All in Interval and principal direction vector Pointing towards the center of the shoulder joint. Ultimately, the geometric basis vectors will be... With this energy characteristic mode Establish a connection, in which This represents the center of the corresponding character's skeletal joint that drives the deformation of this area. In this example, it specifically refers to the center of the shoulder joint. It is stored in the dictionary as a radial fold primitive, thus realizing the feature binding from perceptual image data to rational physical laws.

[0080] In one possible implementation, combining Figure 2 Based on the spatial gradient of the dynamic energy distribution field, generating secondary detailed fold data on the basis of the macroscopic fold geometry data includes:

[0081] Calculate the gradient vector and gradient magnitude of the dynamic energy distribution field at each vertex of the two-dimensional polygonal mesh;

[0082] Vertices whose gradient magnitude exceeds a preset detail generation threshold used to control detail density are selected to form detail generation regions;

[0083] Based on the gradient vector direction at the detailed generation region, a high-frequency perturbation signal with frequency and amplitude related to the gradient amplitude is generated.

[0084] The high-frequency perturbation signal is converted into microscopic geometric displacement to generate secondary detail wrinkle data.

[0085] In some implementations, secondary detail wrinkle data is generated based on the macroscopic wrinkle geometry data according to the spatial gradient of the dynamic energy distribution field. This adds high-frequency, subtle surface undulations to the already formed macroscopic wrinkles to simulate the complex small wrinkles generated by real fabrics in stress concentration areas, thereby enhancing visual realism. The process is initiated by calculating the gradient vector and gradient magnitude of the dynamic energy distribution field at each vertex of a two-dimensional polygonal mesh. For each mesh vertex, the spatial rate of change of the energy field at that point, i.e., the gradient, is calculated using the finite difference method by analyzing its own energy vector and that of its neighboring vertices. The output includes a gradient vector and gradient magnitude for each vertex. The gradient vector points in the direction of the fastest energy growth, and the gradient magnitude is a scalar value representing the degree of change in the energy field at that point. Regions with sufficiently drastic energy changes are identified, necessitating the generation of detail wrinkles, thus avoiding unnecessary calculations in smooth areas. Vertices with gradient magnitudes exceeding a preset detail generation threshold used to control detail density are selected. This detail generation threshold is a globally adjustable normalization parameter, typically set between 0.6 and 0.85. A vertex is included in the set of detail generation regions only when its gradient magnitude exceeds a certain threshold, serving as a candidate point for subsequent detail generation. Small perturbation signals with directionality, frequency, and amplitude consistent with the physical state are created for the vertices of the detail regions. High-frequency perturbation signals with frequencies and amplitudes related to the gradient magnitude are generated based on the gradient vector direction at the detail generation region. For each vertex located within the detail generation region, a geometric displacement is generated based on its gradient information; this process is represented by the following function:

[0086] ;in, This represents the final calculated microscopic geometric displacement vector. It is the gradient magnitude at that vertex. It is the direction of the gradient vector, and both are calculated in the first step. It is the world coordinate position vector of that vertex. It is the normal vector of the vertex, ensuring that displacement occurs along the surface normal. and These are global amplitude and frequency control coefficients used to artistically adjust the intensity and density of detailed wrinkles. The core of this function lies in the gradient magnitude. Simultaneously modulating the amplitude and frequency of the displacement, the larger the gradient, the deeper and denser the resulting detail wrinkles. The direction of the displacement is determined by the gradient vector. The decision is made that the ripple direction of the perturbation signal is typically orthogonal to the gradient direction to simulate fine wrinkles perpendicular to the stretching or compressing direction. The calculated abstract perturbation signal is then transformed into actual geometric data. This involves converting the high-frequency perturbation signal, i.e., the microscopic geometric displacements of each vertex, into... This data is applied directly to the corresponding vertices, or more efficiently, its information is encoded as perturbations in the normal map, thereby generating the final secondary detail wrinkle data. This data is separate from the macroscopic wrinkle geometry data and is specifically designed to describe the microscopic details of the garment surface.

[0087] For example, the gradient magnitude at a certain vertex of the minute stress fluctuations at the edge of a macroscopic fold can be calculated using the finite difference method. The gradient vector direction is 0.9. for Since the amplitude exceeded the set detail generation threshold of 0.7, this point was marked as a detail generation region. Subsequently, a perturbation function was applied. Calculate micro-displacements and set amplitude coefficients. The frequency coefficient is 0.02. The value is 10.0, and the normal at that point is... for Vertex coordinates Projection along the gradient direction The value is 1.2. Substituting the numerical values, we obtain the amplitude term as follows: The phase term is In radians, the displacement vector Through this calculation, high-frequency fluctuations with amplitude and energy change rate are generated on the originally flat macroscopic folds, thereby simulating the unique fine texture of real fabrics through microscopic geometric displacement without increasing the macroscopic geometric burden.

[0088] In one possible implementation, combining Figure 2 The temporal consistency fusion and rendering processing of the macroscopic wrinkle geometry data and the secondary detail wrinkle data includes:

[0089] The macroscopic wrinkle geometry data and the secondary detailed wrinkle data of the current frame are fused together to obtain the total wrinkle geometry data of the current frame.

[0090] Obtain the total fold geometry data generated after processing the previous frame;

[0091] The total folded geometric data of the current frame and the total folded geometric data of the previous frame are processed by a temporal convolution kernel to perform smoothing filtering to obtain smoothed total folded geometric data.

[0092] The smoothed total fold geometry data is converted into normal map information and combined with the basic clothing texture for rendering, outputting a dynamic clothing fold image.

[0093] In some implementations, temporal consistency fusion and rendering processing is performed on macroscopic wrinkle geometry data and secondary detail wrinkle data. This effectively merges the separately calculated wrinkle data of the two scales and eliminates inter-frame jumps through temporal filtering, ultimately generating a visually smooth and natural dynamic wrinkle image. The macroscopic wrinkle geometry data and secondary detail wrinkle data of the current frame are fused to obtain the total wrinkle geometry data of the current frame. This is achieved by vector summing of the microscopic geometric displacements representing secondary detail wrinkle data and the geometric displacements representing macroscopic wrinkle geometry data, or by standardizing and superimposing the perturbation normals generated by both in normal space. This results in a complete geometric deviation containing wrinkle information at all scales at each mesh vertex, i.e., the total wrinkle geometry data of the current frame. Historical data is acquired to provide a reference for achieving temporal smoothing. The total wrinkle geometry data generated after processing the previous frame is retrieved from the cache. This data represents the final geometric state used for rendering after complete fusion and smoothing processing of the previous rendering frame, representing the stable form of the wrinkles at the previous moment. In the first frame of startup, since there is no data from the previous frame, the data is initialized to a zero vector or a neutral geometric state. Unnatural flickering or jumps in the wrinkles between adjacent frames caused by abrupt changes in the input motion data or computational errors are eliminated. The total wrinkle geometry data of the current frame and the processed total wrinkle geometry data of the previous frame are smoothed using a temporal convolution kernel. In engineering implementations, this temporal convolution kernel is typically a simple weighted average or exponential moving average filter, and its operation is as follows: ;in, The current frame The final output is the smoothed total wrinkle geometry data. The current frame Total wrinkle geometry data obtained by direct fusion without smoothing. The previous frame The final output result. These are the weight coefficients of the temporal convolution kernel, floating-point numbers between 0 and 1, typically ranging from 0.1 to 0.4. Smaller... The higher value produces a stronger smoothing effect, making the wrinkle changes more inertial, but with a slight delay in response. Through this filtering operation, the wrinkle shape of the current frame will partially inherit from the previous frame, thus ensuring the temporal consistency of dynamic changes. The calculated pure geometric data is converted into a format that the rendering engine can efficiently utilize, completing the final image output. The smoothed total wrinkle geometry data is converted into normal map information. Based on the final displacement of each vertex after smoothing, the surface normal direction of its mesh cell is recalculated, and these new normal vectors are encoded into a two-dimensional texture, namely the normal map. During the rendering stage, the graphics shader combines this dynamically generated normal map with the basic clothing texture of the garment itself, such as color and pattern. Lighting calculations are performed based on the fine surface directions provided by the normal map, rendering a realistic dynamic clothing wrinkle image with rich lighting variations on the flat surface of the clothing model.

[0094] For example, consider the folds of a long skirt while walking. The macroscopic displacement vector generated in the current frame is merged with the secondary detail displacement vector. Let the fused total fold geometry at a certain vertex be... The displacement is 2.0. This retrieves the smoothed displacement data of this vertex in the final output of the previous frame. The value is 1.5. Applying the temporal convolution kernel formula... Perform filtering and set smoothing weight coefficients. The value is 0.3. Substituting this into the numerical calculation, the final smooth displacement of the current frame is... The smoothed displacement is converted into color information in the normal map, and then the shader is used to blend the lighting with the original fabric texture. This results in a smooth, physically inertial, and flicker-free visual transition effect for the skirt folds as they change from 1.5 to 2.0.

[0095] In one possible implementation, combining Figure 2 The method further includes:

[0096] Real-time monitoring of the character's skeletal motion data, and calculation of a global motion intensity index that quantifies the severity of its changes;

[0097] The global motion intensity index is compared with a stability threshold used to distinguish motion states;

[0098] When the global motion intensity index exceeds the stability threshold, the smoothing intensity of the temporal convolution kernel is adaptively enhanced to maintain the visual stability of the wrinkles.

[0099] In some implementations, an adaptive control mechanism is established to dynamically adjust the intensity of temporal smoothing based on the intensity of the character's movement. This suppresses visual noise during high-speed motion and maintains responsiveness to details during static or slow motion. A global motion intensity index, quantifying the intensity of changes, is calculated by monitoring the character's skeletal motion data in real time. In each frame, a predefined set of key character skeletal joints, such as the pelvis, spine, shoulders, and knees, is traversed, and their angular velocities relative to the previous frame are calculated. The global motion intensity index is obtained by a weighted sum of these angular velocities, as shown below: ;in, It is a global motion intensity index for the current frame, a dimensionless scalar value. It is a joint The magnitude of the angular velocity is obtained by calculating the rate of change of its rotational quaternion. It corresponds to the joint. The weighting coefficients are typically higher for core joints such as the pelvis and spine, ranging from 0.5 to 1.0, while the weights for distal joints are lower, ranging from 0.1 to 0.5. Based on the intensity of movement, the character's movement state is divided into "stable" and "vigorous" modes, providing clear triggering conditions for subsequent adaptive adjustments. The global motion intensity index is compared with a stability threshold used to distinguish movement states. This stability threshold is a pre-tuned floating-point number, for example, set to 50.0, with the equivalent aggregate value in degrees per second, used to define stillness or slow walking versus vigorous movements such as running and jumping. When the global motion intensity index... When the motion intensity exceeds this stable threshold, the current character is determined to be in a state of intense motion. Adaptive adjustments are then performed, modifying the parameters of the temporal filter to handle different motion states and thus optimize the final visual effect. When the global motion intensity index exceeds the stable threshold, the smoothing strength of the temporal convolution kernel is adaptively increased. The weight coefficients used for temporal smoothing filtering are reduced. The value of . For example, in a steady state, The value may be 0.4 to ensure that the folds respond quickly to subtle movements; once in a state of vigorous movement, they will automatically... The value decreases to a smaller value, such as 0.1. According to the smoothing function formula, a smaller value... The value represents the calculation result for the current frame. The weighting is reduced, while the stable result of the previous frame is maintained. The weighting is increased. This enhanced smoothing effect effectively suppresses wrinkle jumps and flickering caused by high-speed motion and rapid data changes, thus maintaining the visual stability of dynamic clothing folds. When the motion intensity falls below a threshold, the weighting coefficient... It will revert to its normal value. For example... Figure 4As shown, when the global motion intensity index exceeds the stability threshold, the weighting coefficient... Non-linear drops occur, thereby increasing the weight of historical frames to ensure visual stability during periods of intense motion.

[0100] For example, when suddenly transitioning from a stationary state to rapid running, a global motion intensity index is calculated by monitoring skeletal data. Assuming the weight of the core joint pelvis Its angular velocity is 1.0. The speed reaches 80.0 at the instant of running, and the calculation is as follows: It is compared with the preset stability threshold of 50.0, because... The system determines that the current state is one of intense motion. An adaptive adjustment mechanism is triggered, adjusting the weights of the temporal convolution kernel. The value was reduced from 0.4 in the steady state to 0.1, and then substituted into the smoothing formula. Perform calculations; if the current frame's fused displacement is... The displacement is 3.0, the same as the displacement in the previous frame. If the value is 1.0, then the final output displacement is Compared to the unadjusted state The calculated value of 1.8 indicates that this adaptive mechanism effectively suppresses visual noise and flickering that may occur during high-speed motion by preserving the stability of the previous frame to a greater extent, thus ensuring the stability of the visual image.

[0101] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0102] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A dynamic fitting method for 2D garment creases based on topological interpolation, characterized in that, The method includes: Obtain the 2D pattern polygon mesh of the garment to be processed and the skeletal motion data of the character driving the garment; The two-dimensional polygonal mesh is dynamically physical parameterized, and each cell in the mesh is assigned a set of dynamic physical properties including basic deformation stiffness and energy conduction rate factor. Based on the character skeleton motion data, dynamic physical attribute set, and the topological connectivity of the mesh, a dynamic energy distribution field covering the entire clothing area is calculated using a topological interpolation algorithm. This calculation includes: calculating virtual forces applied to each vertex of the two-dimensional polygonal mesh based on joint rotation information in the character skeleton motion data; constructing a topological relationship model describing the energy conduction relationship between mesh cells based on the energy conduction rate factor in the dynamic physical attribute set; and simulating the energy conduction and superposition process on the two-dimensional polygonal mesh based on the virtual forces and the topological relationship model, thereby calculating a dynamic energy distribution field composed of energy intensity and energy direction vectors. The nonlinear synthesis of fold morphology based on the dynamic energy distribution field involves comparing the energy intensity of the dynamic energy distribution field with a fold generation threshold dynamically determined according to material properties. Fold geometric patterns are activated and synthesized from a pre-constructed fold morphology dictionary built through manifold learning to generate macroscopic fold geometric data. The nonlinear synthesis of fold morphology based on the dynamic energy distribution field includes: performing nonlinear topological interpolation on the dynamic energy distribution field over time based on the topological relationship model to obtain an energy field evolution sequence containing energy intensity evolution and direction vector evolution; identifying a set of mesh vertices in the energy field evolution sequence whose energy intensity exceeds the fold generation threshold to form fold trigger points; matching several candidate fold primitives from the pre-constructed fold morphology dictionary based on the energy intensity value and energy direction vector at the fold trigger points; and performing geometric transformation and intensity modulation on the matched candidate fold primitives, then superimposing them onto the two-dimensional polygonal mesh to generate macroscopic fold geometric data. Based on the spatial gradient of the dynamic energy distribution field, secondary detailed fold data are generated on the basis of the macroscopic fold geometry data; The macroscopic fold geometric data and the secondary detailed fold data are subjected to temporal consistency fusion and rendering processing to output a dynamic clothing fold image.

2. The method for dynamic fitting of 2D clothing creases based on topological interpolation according to claim 1, characterized in that, Dynamic physical parameterization of the two-dimensional polygonal mesh includes: Configure each cell of the two-dimensional polygonal mesh with an initial set of physical properties containing standardized parameter values; Obtain the fabric type identifier used to specify the type of garment fabric to be processed; Based on the fabric type identifier, retrieve the corresponding reference physical parameter range from a database that stores the physical properties of several real fabrics. Based on the reference physical parameter range, the parameters in the initial physical attribute set are mapped and adjusted to generate a dynamic physical attribute set.

3. The method for dynamic fitting of 2D clothing creases based on topological interpolation according to claim 1, characterized in that, The wrinkle generation threshold dynamically determined based on material properties includes: Obtain the basic deformation stiffness value of each mesh element from the dynamic physical property set; The basic deformation stiffness value is mapped to an energy threshold through a preset functional relationship; The energy threshold is used as the wrinkle generation threshold, wherein a high foundation deformation stiffness value is mapped to a high wrinkle generation threshold, and a low foundation deformation stiffness value is mapped to a low wrinkle generation threshold.

4. The method for dynamic fitting of 2D clothing creases based on topological interpolation according to claim 1, characterized in that, The fold morphology dictionary includes: Acquire images of the sample garment under different motion states, and construct a two-dimensional fold sample set; Manifold learning dimensionality reduction is performed on the two-dimensional fold sample set to extract the low-dimensional fold feature manifold; Basis decomposition is performed on the wrinkled feature manifold to identify several basis vectors representing typical wrinkling patterns; The aforementioned basis vectors and their corresponding energy accumulation patterns are associated and stored in the library as fold primitives to construct a fold morphology dictionary.

5. The method for dynamic fitting of 2D clothing creases based on topological interpolation according to claim 1, characterized in that, Based on the spatial gradient of the dynamic energy distribution field, secondary detailed fold data is generated from the macroscopic fold geometry data, including: Calculate the gradient vector and gradient magnitude of the dynamic energy distribution field at each vertex of the two-dimensional polygonal mesh; Vertices whose gradient magnitude exceeds a preset detail generation threshold used to control detail density are selected to form detail generation regions; Based on the gradient vector direction at the detailed generation region, a high-frequency perturbation signal with frequency and amplitude related to the gradient amplitude is generated. The high-frequency perturbation signal is converted into microscopic geometric displacement to generate secondary detail wrinkle data.

6. The method for dynamic fitting of 2D clothing creases based on topological interpolation according to claim 1, characterized in that, The temporal consistency fusion and rendering processing of the macroscopic wrinkle geometry data and the secondary detailed wrinkle data includes: The macroscopic wrinkle geometry data and the secondary detailed wrinkle data of the current frame are fused together to obtain the total wrinkle geometry data of the current frame. Obtain the total fold geometry data generated after processing the previous frame; The total folded geometric data of the current frame and the total folded geometric data of the previous frame are processed by a temporal convolution kernel to perform smoothing filtering to obtain smoothed total folded geometric data. The smoothed total fold geometry data is converted into normal map information and combined with the basic clothing texture for rendering, outputting a dynamic clothing fold image.

7. The method for dynamic fitting of 2D clothing creases based on topological interpolation according to claim 6, characterized in that, The method further includes: Real-time monitoring of the character's skeletal motion data, and calculation of a global motion intensity index that quantifies the severity of its changes; The global motion intensity index is compared with a stability threshold used to distinguish motion states; When the global motion intensity index exceeds the stability threshold, the smoothing intensity of the temporal convolution kernel is adaptively enhanced to maintain the visual stability of the wrinkles.

8. A 2D clothing crease dynamic fitting system based on topological interpolation, characterized in that, The system is used for a 2D garment crease dynamic fitting method based on topological interpolation as described in any one of claims 1-7, the system comprising: The data acquisition module is used to acquire the two-dimensional polygonal mesh of the garment to be processed and the skeletal motion data of the character driving the garment; The physical parameterization module is used to dynamically physical parameterize the two-dimensional polygonal mesh, assigning each cell in the mesh a set of dynamic physical properties including basic deformation stiffness and energy conduction rate factor. The energy field calculation module is used to calculate the dynamic energy distribution field covering the entire clothing area based on the character skeleton motion data, dynamic physical attribute set, and the topological connection relationship of the mesh, using a topological interpolation algorithm. The calculation of the dynamic energy distribution field covering the entire clothing area using the topological interpolation algorithm includes: calculating virtual forces applied to each vertex of the two-dimensional polygonal mesh based on joint rotation information in the character skeleton motion data; constructing a topological relationship model describing the energy conduction relationship between mesh cells based on the energy conduction rate factor in the dynamic physical attribute set; and simulating the energy conduction and superposition process on the two-dimensional polygonal mesh based on the virtual forces and the topological relationship model, to calculate the dynamic energy distribution field composed of energy intensity and energy direction vectors. A macroscopic wrinkle synthesis module is used for nonlinear synthesis of wrinkle morphology based on the dynamic energy distribution field. This involves comparing the energy intensity of the dynamic energy distribution field with a wrinkle generation threshold dynamically determined based on material properties, activating and synthesizing wrinkle geometric patterns from a pre-constructed wrinkle morphology dictionary built through manifold learning, and generating macroscopic wrinkle geometric data. The nonlinear synthesis of wrinkle morphology based on the dynamic energy distribution field includes: performing nonlinear topological interpolation on the dynamic energy distribution field over time based on the topological relationship model to obtain an energy field evolution sequence containing energy intensity evolution and direction vector evolution; identifying a set of mesh vertices in the energy field evolution sequence whose energy intensity exceeds the wrinkle generation threshold, forming wrinkle trigger points; matching several candidate wrinkle primitives from the pre-constructed wrinkle morphology dictionary based on the energy intensity value and energy direction vector at the wrinkle trigger points; and performing geometric transformation and intensity modulation on the matched candidate wrinkle primitives, then superimposing them onto the two-dimensional polygonal mesh to generate macroscopic wrinkle geometric data. The detail wrinkle generation module is used to generate secondary detail wrinkle data based on the macroscopic wrinkle geometry data according to the spatial gradient of the dynamic energy distribution field. The fusion rendering module is used to perform temporal consistency fusion and rendering processing on the macroscopic fold geometric data and the secondary detailed fold data, and output dynamic clothing fold images.

Images