A method and system for geometric simulation and high-fidelity rendering of a knitted fabric

Abstract

The application discloses a kind of knitted fabric geometric simulation and high realistic rendering method and system.The method is based on knitted loop type point sequence, generates smooth center line by parametric curve reconstruction, constructs yarn three-dimensional geometry with twist and cross-sectional characteristics;Then construct structured particle hair generation system on yarn surface, with normal vector as the leading, tangential vector and curvature as the disturbance term;Geometric node and force field control are used to accurately optimize particle units on the Blender platform, simulate the natural physical interaction and micro-morphology of fiber.In the rendering level, the application breaks through the traditional geometric coloring limitations by integrating displacement map, particle system and hair BSDF shader, and realizes multi-scale high-physical realistic rendering from macro-morphology to micro-hair through the cooperation of the three.Finally, through geometry simplification and texture baking technology, interactive preview model is generated by importing real-time rendering engine.The application provides a high realistic digital new technical path for the modeling and rendering of knitted fabric.

Description

Technical Field

[0001] This invention relates to the interdisciplinary fields of computer graphics, computer-aided design, and textile engineering. Specifically, it relates to a multi-scale, highly realistic 3D simulation and rendering method for knitted fabrics, from yarn geometry to fiber hairiness. Background Technology

[0002] In the fields of apparel digitization, virtual try-on, and digital textiles, achieving highly realistic 3D simulation of knitted fabrics is a key technological challenge. The visual appearance of knitted fabrics is determined by their macroscopic structure, yarn 3D morphology, and surface fiber hairiness. Existing technologies have significant shortcomings at different levels: at the macroscopic level, mainstream commercial software (such as CLO 3D and Marvelous Designer) uses fabric physics simulation based on triangular meshes, which can effectively simulate the drape and dynamics of the fabric, but cannot represent the interlacing structure and twist details at the yarn level, resulting in renderings that lack the unique three-dimensional texture of knitted fabrics; at the yarn-level modeling level, academic research often uses NURBS curves to construct the yarn centerline and generates smooth tubular geometry through scanning. While these methods can represent yarn direction, the model surface is smooth and monotonous, completely lacking the key microscopic feature that determines the fabric's visual softness, fluffiness, and optical properties—fiber hairs. In terms of hair representation, existing technologies either use completely random particle systems in 3D software to simulate hairs, resulting in chaotic hair direction unrelated to the yarn's main geometry and poor physical realism; or they employ high-precision fiber-level physical simulation, which is computationally expensive and only suitable for offline rendering, failing to meet the needs of interactive applications. Furthermore, there is a technological gap between high-fidelity rendering and real-time rendering. Film-grade hair rendering technology can produce highly realistic hair effects, but its massive computational demands make it difficult to apply to real-time interactive scenarios. Real-time rendering solutions typically sacrifice geometric details, using screen-space techniques such as normal mapping and parallax masking for visual simulation, which easily reveals flaws under close-up observation or dynamic perspectives, resulting in limited fidelity.

[0003] Therefore, there is an urgent need in this field for an innovative method that can accurately model the geometry of yarn, generate structured hairs that conform to physical laws, and render high-quality offline at a cinematic level, providing an innovative technical path for the digital modeling and realistic rendering of knitted fabrics. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, this invention provides a method for yarn-level geometric simulation and high-fidelity rendering of knitted fabrics. This invention aims to construct a highly realistic yarn detail building system, realizing a complete technology from loop shape point input, yarn volume generation, and feather system construction to high-fidelity rendering. With the core objective of achieving a realistic end-to-end presentation of knitted fabrics from yarn geometry to feather details, a high-fidelity rendering system based on Blender is constructed. Employing a data-driven method based on shape points, it establishes a complete technical link from data input to rendering output through curve reconstruction of the loop center path, comprehensive generation of yarn geometric volume, and structured control of the feather particle system.

[0005] This invention provides a geometric simulation and high-fidelity rendering of knitted fabrics, specifically implemented through the following steps:

[0006] Step 1: Obtain the sequence of shape points and reconstruct the centerline by fitting the shape points;

[0007] Step 2: Generate the yarn mesh volume containing twist;

[0008] Step 3: Introduce the Frenet framework to generate tangential-normal fields, and construct the twist angle function and the spatial path at the yarn fiber level;

[0009] Step 4: Construct the corrected Bishop frame, map the original normal of the mesh surface to the smooth vector field of the Bishop frame, and calculate the normal vector of each surface sampling point.

[0010] Step 5: Export the local normal, tangential and curvature to construct the composite direction field, and calculate the initial direction of the feathers;

[0011] Step 6: Constrain the hairs to achieve structured growth along the surface with the yarn in an inclined direction;

[0012] Step 7: Construct a multi-layered feather structure, adjust the feather density and length, and present a three-layer overlapping visual effect from the down layer to the edge;

[0013] Step 8: Integrate the hair bidirectional scattering distribution function, multi-scale normal mapping and volume scattering algorithm to perform high-fidelity rendering and generate a high-realism knitted fabric simulation model.

[0014] Furthermore, in step 1, firstly, the sequence of shape value points of the knitted loop is obtained, and a stable loop model is constructed using the geometric modeling method of parametric curves.

[0015] Next, construct a controllable coil centerline. This enables adjustable, stable, and renderable geometry of knitted yarns, where parameter t controls the position of the curve on the path.

[0016] Finally, in the coil reconstruction stage, based on the yarn spatial orientation and coil geometric characteristics, a smooth periodic curve function with multiple parameters is used to describe the three-dimensional shape of a single coil; the parameterized curve function expression method for reconstructing the center line of the fitting value point is shown in Equation (1).

[0017] in, With the center line as the reference point, 'a' is a parameter that controls the roundness of the coil, 'h' determines the height of the coil along the tissue direction, and 'd' determines the spatial tension of the coil in the front-to-back direction.

[0018] Furthermore, the tangential-normal field includes ,in It is a tangent. and They are normal and secondary normal; The principle of the twist angle function is shown in equation (2):

[0019] The calculation principle of the spatial path at the yarn fiber level is shown in equation (3):

[0020]

[0021] In the formula, Centerline This is an extended parameter for the yarn radius. It is a constant representing the initial angle of the yarn fiber at the starting point. The yarn twist is represented by parameter t, which controls the position of the curve on the path.

[0022] Furthermore, the revised Bishop frame is ,in It is a tangent. A normal vector is a vector that is perpendicular to each other in the normal plane and evolves smoothly with the curve. The calculation principle is shown in equation (4):

[0023]

[0024] Where s is the arc length of the curve, corresponding to the arc length of the parameter value t. It is the rotation angle around the tangent. Further, in step 5, firstly, from the center line... Derivation of local normal With tangent ;

[0025] Subsequently, during the particle emission phase Main direction, with As a slight offset direction, it makes the hairs exhibit the characteristic of naturally tilting along the yarn;

[0026] Then, the curvature is superimposed. The directional distribution of the feathers is perturbed to make the feathers in the coil bending area exhibit a more natural diffusion trend; among which, the curvature From the center line It is obtained by calculating the derivative, specifically... Defined as the magnitude of the derivative of the tangent vector, its calculation principle is shown in equation (5):

[0027]

[0028] in, That is Next, calculate the second normal vector. In the Frenet frame, By tangent vector and principal normal vector The definition of the cross product, and in the revised Bishop frame In the middle, still satisfied The calculation principle is shown in equation (6):

[0029]

[0030] Right now,

[0031] Finally, regarding particle path control, in order to maintain the overall directional constraint while exhibiting flexible fiber characteristics, the final hair generation path, i.e., the initial growth direction of the hair, is obtained. The calculation principle is shown in equation (7):

[0032]

[0033] in, The normal dominance factor ensures that hair grows along the normal direction of the yarn surface. This is the tangential offset coefficient, used to control the degree of inclination of the hairs along the yarn axis. The curvature perturbation coefficient adjusts the diffusion intensity of hairs in the bending region.

[0034] Furthermore, in step 6, firstly, the growth direction of the hair is dominated by the yarn cross-section normal, thereby determining the reference growth axis of each hair in three-dimensional space; then, the tilt angle of the hair is finely adjusted by linear interpolation, thereby simulating the directional or helical tilt characteristics formed by the fiber during twisting and spinning.

[0035] Finally, fine-tuning is performed based on tangentiality and curvature to obtain the three-dimensional path of a single hair. The calculation principle is shown in equation (8):

[0036]

[0037] in, It is the length parameter of the feathers themselves, defined in the interval ∈[0, 1] indicates from the root of the hair. =0 to the tip =1 normalized position; u(t, This represents the mapping of parameter t to a specific yarn position and the position of a specific length of hairiness. The displacement vector of the disturbance superimposed at the location.

[0038] Furthermore, in step 7, firstly, the density, length, and softness of the feathers are manually adjusted according to the yarn thickness, local stress, and different coil contact areas; then, from the perspective of visual performance and function, the feather system is layered into a three-layer overlapping structure of down layer, feather layer, and free feathers; finally, in the real-time interaction stage, the natural movement and physical interaction of the feathers are simulated through force field control and collision detection.

[0039] Furthermore, the specific implementation method of step 8 is as follows:

[0040] First, a microscopic optical representation layer is constructed by integrating physical displacement maps and hair bidirectional scattering distribution functions. The grayscale height information of the displacement maps is used to replace the geometry of the yarn substrate, accurately shaping the spiral twisting texture between individual strands and the macroscopic surface unevenness. Simultaneously, a hair bidirectional scattering distribution function shader is configured, using its dual-mirror reflection model to simulate the reflection, refraction, and internal reflection behavior of light on the surface of microfibers. A volume scattering algorithm is introduced to reproduce the unique semi-transparent edges and soft halo effect of knitted fabrics by calculating the mean free path and light energy transfer law inside the hair layer.

[0041] Subsequently, the multi-layered hair structure was sampled for spatial position and extracted for orientation field, baked into a shell texture with multi-level depth, and a visual substitution of tens of thousands of fibers was achieved by using a limited geometric patch in combination with a transparency mask to obtain a knitted fabric simulation model; at the same time, the knitted fabric simulation model was optimized and compressed by geometric mesh topology.

[0042] Finally, the optimized knitted fabric simulation model is exported using a standard model exchange format, and the fur material is reconstructed. A custom shader is used to approximate the anisotropic reflection of the knitted fabric simulation model. Combined with real-time global illumination and shadow algorithms, the knitted fabric simulation model can maintain stable visual fidelity and physical texture under different dynamic perspectives.

[0043] The present invention also provides a geometric simulation and high-fidelity rendering system for knitted fabrics, including a processor and a memory. The memory is used to store program instructions, and the processor is used to call the program instructions in the memory to execute a geometric simulation and high-fidelity rendering method for knitted fabrics as described in the above technical solution.

[0044] The present invention also provides a computer-readable storage medium, including a readable storage medium on which a computer program is stored, wherein when the computer program is executed, a geometric simulation and high-realism rendering method for knitted fabrics as described in the above technical solution is implemented.

[0045] Through the above steps, this invention successfully establishes a complete technical chain from coil-level structural modeling to high-performance real-time display. It not only achieves cinematic-level visual reproduction of knitted fabrics at the microscopic scale but also ensures the smooth operation and high-quality application of high-fidelity assets in interactive scenarios. This invention overcomes the limitations of traditional geometric shading by integrating displacement mapping, particle systems, and hair BSDF shaders: displacement mapping realistically alters the base geometry based on grayscale information to reflect the twisting of strands; particle systems generate controlled hair textures to simulate fiber details; and the hair BSDF model accurately calculates the anisotropic reflection and scattering of light. These three elements work together to achieve multi-scale, highly physically realistic rendering from macroscopic morphology to microscopic hair texture. Finally, through geometric simplification and texture baking techniques, the model is imported into a real-time rendering engine to generate an interactive preview model. This invention systematically introduces film-level hair rendering technology into textile simulation, improving the accuracy of microscopic optical models and providing a highly realistic digital new technology path for the modeling and rendering of knitted fabrics. Attached Figure Description

[0046] Figure 1 This is a flowchart illustrating the overall method of an embodiment of the present invention.

[0047] Figure 2 The result of reconstructing the centerline for the fitted value points.

[0048] Figure 3 Effect diagram of yarn volume with twist content.

[0049] Figure 4 The image shows the effect of generating the feather orientation field under the composite orientation field.

[0050] Figure 5 A schematic diagram illustrating the principle of generating structural hairs along the surface with a slanted, yarn-oriented orientation.

[0051] Figure 6 The images show the multi-layered feather structure, with (a) and (b) showing details of different parts.

[0052] Figure 7 This image shows the effect of integrating BSDF for hair and multi-scale normal mapping.

[0053] Figure 8 The image shows a highly realistic rendering of the yarn, where (a) and (b) are the front and back sides of the fabric, respectively. Detailed Implementation

[0054] The technical solution of this invention can be implemented by those skilled in the art using computer software technology.

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be noted that the following embodiments are only for explaining the invention and not for limiting it. The software platforms used in implementing this invention (such as Blender, Unity) are examples and do not constitute a limitation on the scope of protection. This embodiment uses the following specific steps to implement high-fidelity simulation in the open-source 3D software Blender:

[0056] Combined with appendix Figure 1 This invention provides a real-time rendering method for knitted fabrics based on deferred shading, which specifically includes the following steps:

[0057] Step 1: Obtain the sequence of shape points and reconstruct the center line by fitting the shape points.

[0058] Step 2: Generate the yarn mesh volume containing twist.

[0059] Step 3: Introduce the Frenet framework to generate tangential-normal fields, and construct the twist angle function and the spatial path at the yarn fiber level;

[0060] Step 4: Construct the corrected Bishop frame, map the original normal of the mesh surface to the smooth vector field of the Bishop frame, and calculate the normal vector of each surface sampling point.

[0061] Step 5: Export the local normal, tangential and curvature to construct a composite direction field, and calculate the initial direction of the feathers.

[0062] Step 6: Constrain the hairs to achieve structured growth along the surface with the yarn inclined.

[0063] Step 7: Construct a multi-layered feather structure, adjusting the density and length according to stress to present a three-layer overlapping visual effect from the down layer to the edge.

[0064] Step 8: Integrate the hair bidirectional scattering distribution function (BSDF), multi-scale normal mapping, and volume scattering algorithm for high-fidelity rendering. After geometric simplification and texture baking, generate a high-fidelity knitted fabric simulation model.

[0065] The following examples illustrate the processing procedure for each step: The examples are based on a model file of a knitted fabric, and the method of the present invention is tested.

[0066] In step 1, the example uses the spatial coordinates of the shape points in the model file of the knitted fabric as a basis to reconstruct the center line by fitting the shape points. The specific method is as follows:

[0067] First, obtain the sequence of shape value points of the knitted loops, and then construct a stable loop model using the geometric modeling method of parametric curves.

[0068] Next, construct a controllable coil centerline. This enables adjustable, stable, and renderable geometry of knitted yarns. The parameter 't' controls the curve's position on the path. .

[0069] Finally, in the coil reconstruction stage, based on the yarn spatial orientation and coil geometric characteristics, a smooth periodic curve function with multiple parameters is used to describe the three-dimensional shape of a single coil. The parameterized curve function expression method for reconstructing the center line of the fitted value point is shown in Equation (1).

[0070]

[0071] Where 'a' controls the roundness of the coil, 'h' determines the height of the coil along the tissue direction, and 'd' determines the spatial tension of the coil in the front-to-back direction, i.e., the so-called "out-of-plane bending". In the embodiment, 'a' is 0.2, 'h' is 0.5, and 'd' is 0.1.

[0072] Through the above steps, stable reconstruction of the yarn geometry model and continuous controllability of the normal field are achieved. Fitting from discrete value points to a parameterized centerline ensures the adjustability of the coil shape. In this embodiment, a schematic diagram of the centerline reconstruction from fitted value points is shown in the attached diagram. Figure 2 As shown.

[0073] In step 2, a mesh volume containing twisted yarn is generated, which is to create a mesh volume shell for the center line in step 1. The specific method is as follows:

[0074] First, in Blender's 3D creation suite, yarn volume is generated using geometry nodes. Then, the mesh normal node within the geometry nodes is used to obtain the topological normal vectors of the yarn volume model in 3D space. Since the yarn is formed by the centerline... The pipe structure generated by sweeping the profile curve has its normal vector initially defined as a unit vector perpendicular to the centerline tangent and pointing towards the section boundary.

[0075] Subsequently, the physical parameters of the yarn are mapped to the geometry in a programmed manner, incorporating features such as twist and localized torsion. In this embodiment, a yarn ply of 4 is used.

[0076] Finally, using the yarn radius r as an extension parameter, the centerline... The structure is transformed into a volumetric line structure. Using the centerline as a guide, a three-dimensional volumetric model of the yarn is generated by combining the cross-sectional curves. In this embodiment, the radius r is 0.05, and the yarn twist... It is 5.

[0077] Through the above steps, yarn twist and yarn volume control were achieved, laying a stable geometric foundation for subsequent mechanical simulation and optical rendering of yarn hairiness. An example is attached, showing the effect of yarn volume with twist. Figure 3 As shown.

[0078] In step 3, the Frenet framework is introduced to generate the tangential-normal field, construct the twist angle function and the spatial path at the yarn fiber level. The specific method is as follows:

[0079] First, to ensure that the yarn cross-section does not change abruptly, this invention first employs the Frenet (parallel transport) framework to generate a smooth tangential-normal field. .in It is a tangent. and These are the normal and the secondary normal.

[0080] Subsequently, considering that the Frenet framework is prone to torsional singularities in regions of rapid curvature change, an angle correction term was introduced. The twist angle function is constructed to make the twist change more uniform. It is the key angle function that determines the winding position of the yarn fibers. Its calculation principle is shown in Equation (2):

[0081]

[0082] Next, the spatial path at the yarn fiber level, i.e. the initial yarn path, is constructed. The calculation principle is shown in equation (3):

[0083]

[0084] In the formula, This is an extended parameter for the yarn radius. It is a constant representing the initial angle of the yarn fiber at the starting point (t=0 or any point). This refers to yarn twist. This method not only achieves continuous control of yarn twist and fiber winding period, but also ensures initial constraint on the direction of hair generation with the surface normal as the principal axis during the particle emission stage. In the embodiment, yarn twist... Set to 5 to balance the tightness and softness of the yarn and maintain a medium twisting rate.

[0085] Through the above steps, continuous control of yarn twist and fiber winding period was achieved. At the same time, the torsional singularity of the Frenet frame in the curvature change region was effectively suppressed, ensuring the smoothness and physical consistency of the normal vector field, and laying a stable geometric foundation for subsequent mechanical simulation and optical rendering of hair.

[0086] Step 4: Construct the corrected Bishop frame, map the original normal of the mesh surface to the smooth vector field of the Bishop frame, and calculate the normal vector of each surface sampling point.

[0087] First, the parametric centerline is converted using the "Curve to Mesh" function. The mesh is transformed into a tubular geometry with thickness. During this process, the capture attribute mechanism in the geometric nodes is used to lock the relative position of each vertex in the local coordinate system at the instant the mesh is generated. Since the yarn surface is composed of a cross-sectional circle rotating around the centerline, the initial normal vector of each vertex of the mesh points radially away from the centerline by default, providing the original directional reference for subsequent hair growth.

[0088] Next, in order to eliminate the interference of geometric topological deformation on the feather direction, this invention implements a modified Bishop frame by accessing step 3. , They are vectors that are perpendicular to each other in the normal plane and evolve smoothly with the curve.

[0089] The original normal vector of the mesh surface is mapped onto this smooth vector field. The accurate normal vector of each surface sampling point is calculated, and the calculation principle is shown in Equation (4):

[0090]

[0091] Where s is the arc length of the curve, corresponding to the arc length s of parameter value t. It is the angle of rotation around the tangent.

[0092] Finally, the extracted and corrected surface normals will be... This parameter, known as the main axis, is passed to subsequent particle hair generation. It is calculated by intersecting the normal and the yarn tangent. The vector product is used to construct the local coordinate basis for hair growth, and a weighting factor is introduced in subsequent steps to realize the morphological evolution of hair from a simple "radial outward" to "tilted with the yarn". The extracted normal data is stored as a set of multi-dimensional vector attributes, which are used to guide the anisotropic distribution of the hair BSDF shader in the rendering stage, thereby realizing the realistic scattering of light and shadow on the yarn surface.

[0093] Through the above steps, this invention achieves the precise extraction of surface physical properties from abstract centerline geometry, solving the problems of random hair direction and disconnection from yarn structure in traditional methods, and laying a solid geometric foundation for constructing high-fidelity knitted fabric simulation models.

[0094] In step 5, the local normal, tangential, and curvature are derived to construct a composite direction field, and the initial direction of the feather particles is calculated.

[0095] First, by accessing step 4, from the center curve Derivation of local normal With tangent .

[0096] Subsequently, during the particle emission phase Main direction, with As a slight offset direction, it makes the hairs exhibit the characteristic of naturally tilting along the yarn.

[0097] Then, the curvature is superimposed. The distribution of hair direction is disturbed to make the hair in the coil bending area show a more natural diffusion trend. The greater the curvature of the bend, the greater the random offset of the hair direction, but the whole is still constrained by the main normal direction. This simulates the "micro-spreading" effect of surface fibers caused by local deformation when real yarn is bent under force.

[0098] Among them, curvature From the center curve It is calculated from the derivative (tangent vector, normal vector, etc.), specifically, Defined as the magnitude of the derivative of the tangent vector, its calculation principle is shown in equation (5):

[0099]

[0100] Next, calculate the second normal vector. In the Frenet frame, By tangent vector and principal normal vector The definition of cross product ( = × ), and in the corrected approximate Bishop frame In the middle, it is still satisfied ( The calculation principle is shown in equation (6):

[0101]

[0102] Right now,

[0103] Finally, in terms of particle path control, in order to make the hair feathers exhibit flexible fiber characteristics while maintaining the overall directional constraints, the final hair feather generation path is obtained, and its calculation principle is shown in Equation (7):

[0104]

[0105] in, The normal dominance factor ensures that hair growth primarily occurs along the normal direction of the yarn surface. This is the tangential offset coefficient, used to control the degree of inclination of the hairs along the yarn axis. The curvature perturbation coefficient, combined with curvature In the example, the diffusion intensity of the hairs in the bending area is adjusted. It is 0.8. The values ​​are 0.2 and 0.3 respectively.

[0106] Through the above steps, the main direction, fine-tuning direction, and perturbation factor of the feathers are assigned different weights. The initial growth direction of each hair is calculated by mixing the fibers. This hair system, which exhibits both regularity and natural randomness, better reflects the physical laws of real textile materials in terms of spatial distribution, directional consistency, and dynamic performance. In the embodiment, Figure 4 The image shows the effect of generating the feather orientation field under the composite orientation field.

[0107] In step 6, the hairs are constrained to achieve structured growth along the surface with a slant towards the yarn. The specific method is as follows:

[0108] First, in order to make the hairs have the flexible characteristics of real fibers, their main growth direction is dominated by the normal of the yarn cross section, thereby determining the reference growth axis of each hair in three-dimensional space.

[0109] Normal dominance is determined by coefficients The relative size is controlled. The embodiment sets... , It is 0.2. The value is 0.3, which means that when synthesizing the direction vector, the normal component... The tangential component has the largest weight. and subnormal components Its weight is relatively small.

[0110] Then, the tilt angle of the hairline is fine-tuned by linear interpolation to simulate the unidirectional or helical tilting characteristics formed by the fiber during twisting and spinning. In this embodiment, the number of interpolations for the sub-level hairline is 100.

[0111] Finally, fine-tuning is performed based on tangentiality and curvature. The calculation principle of the three-dimensional path of a single hair is shown in equation (8):

[0112]

[0113] in, This represents the perturbation term, used to make the feathers appear as realistic features such as slight curls, gentle wisps, and natural curves. It is the length parameter of the feathers themselves, defined in the interval ( ∈[0, 1]), indicating from the root of the hair ( =0) to the tip ( The normalized position of u(t, =1). This represents the mapping of parameter t to a specific yarn position and the position of a specific length of hairiness. The superimposed disturbance displacement vector at the location, where the input parameter is (t, The output is a displacement increment (a three-dimensional numerical set). To achieve the above perturbation effect, it can be adjusted through a parameterized interface in specific implementations. In the example, = Function(Amplitude, Frequency, Straightness, Cluster, Shape, Knot Type...), where the knot type is set to "Curl" to simulate the natural bending characteristics of the fiber. The amplitude is set to 0.1mm, which is the amplitude of the main low-frequency noise, corresponding to the influence of ambient airflow (wind field) on the overall oscillation amplitude of the fibers. The frequency is 2.5, which is the frequency of the main low-frequency noise, controlling the fluctuation scale of the wind field effect. The straightness, cluster, and shape parameters are set to 0.1, 1.0, and 0 respectively to further fine-tune the shape and distribution of the curl.

[0114] Through the above steps, the feathers maintain both the geometric plausibility of "attaching to the yarn surface" and the physical realism of "swaying gently in the wind and bending according to shape," providing a microscopic model with both geometric accuracy and dynamic features for subsequent physically based rendering. In the embodiment, Figure 5 A schematic diagram illustrating the principle of hair growth along the surface to achieve a slanted, yarn-oriented formation.

[0115] In step 7, a multi-layered feather structure is constructed, and the density and length are adjusted according to stress to present a three-layer overlapping visual effect from the down layer to the edge. The specific method is as follows:

[0116] First, based on the yarn thickness, local stress, and coil contact area, the density, length, and softness of the hairs are manually adjusted at the geometric nodes. Local stress is derived from simulations of the physical changes in the yarn itself, while the coil contact area is derived from geometric analysis of the fabric's macroscopic or microscopic structure. When constructing the 3D model of the knitted fabric, the interlacing and wrapping relationships between yarns are clearly defined. Thicker yarns have higher hair density, shorter lengths, and lower softness; thinner yarns have lower hair density, increased length, and improved softness. In this embodiment, based on spatial distribution and geometric parameters, the hairs on the yarn surface are divided into "<0.05 mm winding layer," "0.05~0.15 mm core layer hairs," and ">0.15 mm free layer" from the yarn surface, generating particle hairs and particles of different lengths on the yarn volume surface, respectively. In the embodiment, the entangled layer fibers emit high-density short-length particles (<0.05 mm from the surface) with a density of 900 particles / cm³, a length of 0.05 mm, and a softness of 0.8; the core layer fibers emit medium-density medium-length particles (0.05~0.15 mm from the surface) with a density of 1500 particles / cm³, a length of 0.1 mm, and a softness of 0.5; and the free layer fibers emit low-density long-length particles (>0.15 mm from the surface) with a density of 500 particles / cm³, a length of 0.2 mm, and a softness of 0.3, thereby achieving the layered generation of the core layer, entangled layer, and free layer fibers.

[0117] Then, from the perspectives of visual representation and function, the hair system is layered into an extremely short pile layer, a moderate hair layer, and relatively long free hairs, thus visually presenting a three-layer overlapping structure of "core-entanglement-free" on the surface of the real yarn. The pile layer covers the yarn substrate to form a uniform diffuse reflection layer, and the surface hairs are oriented along the tangent of the yarn to enhance the continuity of luster. It can be seen that the hairs form a scattering transition zone at the edge to weaken the hard edge of the outline.

[0118] Finally, in the real-time interaction phase, the natural drifting and physical interaction of feathers can be simulated through force field control (such as wind field, which is usually defined as a vector field in three-dimensional space, where each vector represents the wind direction and intensity at that point. During real-time updates, the feather particles (single feathers) are subjected to the force of the wind field vector, thus affecting their position and direction, achieving a drifting effect. The wind field may change over time to simulate the randomness of natural wind.) and collision detection (colliders set the motion boundaries for the feather particles. When a particle attempts to enter the collider, the system adjusts the particle's motion state by calculating rebound, friction, or adhesion to ensure that the feathers naturally stop or slide when interacting with objects).

[0119] Through the above steps, this layered model makes the hair-like texture in the local contact area appear realistically bent due to spatial compression, greatly enhancing the three-dimensionality of the optical features. Through these steps, this layered model not only restores the multi-scale hair-like topology of real yarn, but also achieves coupled simulation of "morphology-mechanics-optics" through stress-driven dynamic adjustment, providing both geometric and dynamic constraints for the physical realism calculation of subsequent rendering layers. In the embodiment, Figure 6 This is an illustration of a multi-layered feather structure.

[0120] In step 8, the hair bidirectional scattering distribution function (BSDF), multi-scale normal mapping, and volume scattering algorithm are integrated for high-fidelity rendering to generate a high-realism knitted fabric simulation model. The specific method is as follows:

[0121] First, a microscopic optical representation layer is constructed by integrating physical displacement mapping and hair two-way scattering distribution function (Hair BSDF). The grayscale height information of the displacement map (obtained from the initial yarn model) is used to replace the geometry of the yarn substrate, accurately shaping the spiral twisting texture between individual strands and the macroscopic surface unevenness. Simultaneously, a hair BSDF shader is configured, utilizing its dual-mirror reflection model to simulate the reflection, refraction, and internal reflection behavior of light on the surface of microfibers. To enhance the fabric's fluffy texture, a volumetric scattering algorithm is further introduced. By calculating the mean free path and light energy transfer law within the hair layer, the problem of excessive shadow accumulation common in high-density particle systems is effectively solved, reproducing the unique semi-transparent edges and soft halo effect of knitted fabrics. In the embodiments, Figure 7 This image shows the effect of integrating BSDF for hair and multi-scale normal mapping.

[0122] Subsequently, the multi-layered hair structure was sampled spatially and its orientation field extracted, then baked into a shell texture with multi-level depth. A limited number of geometric faces, combined with an opacity mask, were used to visually replace tens of thousands of fibers, resulting in a knitted fabric simulation model. Simultaneously, to meet the performance requirements of the real-time engine, the knitted fabric simulation model underwent geometric mesh topology optimization, and the high-precision microscopic texture was compressed into multi-scale normal maps and ambient occlusion (AO) textures. This significantly reduced the number of polygons while preserving yarn-level detail to the greatest extent possible.

[0123] Finally, the optimized knitted fabric simulation model, baked textures, and related attribute data were losslessly exported using a standard model exchange format, and the physically based rendering (PBR) feather material system was reconstructed in the Unity environment. A custom shader was used to approximate the anisotropic reflections in the offline stage, combined with real-time global illumination and shadow algorithms, ensuring that all knitted fabric simulation models maintained stable visual fidelity and physical texture under different dynamic viewpoints.

[0124] Through the above steps, this invention successfully establishes a complete technical chain from coil-level structure modeling to high-performance real-time display. This not only achieves cinematic-level visual reproduction of knitted fabrics at the microscopic scale but also ensures the smooth operation and high-quality application of high-fidelity assets in interactive scenarios. In the embodiments, Figure 8 This is a rendering of highly realistic yarn.

[0125] Secondly, embodiments of the present invention also provide a geometric simulation and high-fidelity rendering system for knitted fabrics, including a processor and a memory. The memory is used to store program instructions, and the processor is used to call the program instructions in the memory to execute a geometric simulation and high-fidelity rendering method for knitted fabrics as described in the above technical solution.

[0126] Thirdly, embodiments of the present invention also provide a computer-readable storage medium, including a readable storage medium on which a computer program is stored, wherein when the computer program is executed, it implements a geometric simulation and high-realism rendering method for knitted fabrics as described in the above technical solution.

[0127] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for geometric simulation and high-fidelity rendering of knitted fabrics, characterized in that, Includes the following steps: Step 1: Obtain the sequence of shape points and reconstruct the centerline by fitting the shape points; Step 2: Generate the yarn mesh volume containing twist; Step 3: Introduce the Frenet framework to generate tangential-normal fields, and construct the twist angle function and the spatial path at the yarn fiber level; Step 4: Construct the corrected Bishop frame, map the original normal of the mesh surface to the smooth vector field of the Bishop frame, and calculate the normal vector of each surface sampling point. Step 5: Export the local normal, tangential and curvature to construct the composite direction field, and calculate the initial direction of the feathers; Step 6: Constrain the hairs to achieve structured growth along the surface with the yarn in an inclined direction; Step 7: Construct a multi-layered feather structure, adjust the feather density and length, and present a three-layer overlapping visual effect from the down layer to the edge; Step 8: Integrate the hair bidirectional scattering distribution function, multi-scale normal mapping and volume scattering algorithm to perform high-fidelity rendering and generate a high-realism knitted fabric simulation model.

2. The geometric simulation and high-realism rendering method for knitted fabrics as described in claim 1, characterized in that: In step 1, firstly, the sequence of shape value points of the knitted loop is obtained, and a stable loop model is constructed using the geometric modeling method of parametric curves; Next, construct a controllable coil centerline. This enables adjustable, stable, and renderable geometry of knitted yarns, where parameter t controls the position of the curve on the path. Finally, in the coil reconstruction stage, based on the yarn spatial orientation and coil geometric characteristics, a smooth periodic curve function with multiple parameters is used to describe the three-dimensional shape of a single coil; the parameterized curve function expression method for reconstructing the center line of the fitting value point is shown in Equation (1). ； in, With the center line as the reference point, 'a' is a parameter that controls the roundness of the coil, 'h' determines the height of the coil along the tissue direction, and 'd' determines the spatial tension of the coil in the front-to-back direction.

3. The geometric simulation and high-realism rendering method for knitted fabrics as described in claim 1, characterized in that: Tangential-normal fields include ,in It is a tangent. and They are normal and secondary normal; The principle of the twist angle function is shown in equation (2): ； The calculation principle of the spatial path at the yarn fiber level is shown in equation (3): ； In the formula, Centerline This is an extended parameter for the yarn radius. It is a constant representing the initial angle of the yarn fiber at the starting point. The yarn twist is represented by parameter t, which controls the position of the curve on the path.

4. The geometric simulation and high-realism rendering method for knitted fabrics as described in claim 1, characterized in that: The revised Bishop sign is ,in It is a tangent. A normal vector is a vector that is perpendicular to each other in the normal plane and evolves smoothly with the curve. The calculation principle is shown in equation (4): ； Where s is the arc length of the curve, corresponding to the arc length of the parameter value t. It is the angle of rotation around the tangent.

5. The geometric simulation and high-realism rendering method for knitted fabrics as described in claim 4, characterized in that: In step 5, first, from the center line Derivation of local normal With tangent ; Subsequently, during the particle emission phase Main direction, with As a slight offset direction, it makes the hairs exhibit the characteristic of naturally tilting along the yarn; Then, the curvature is superimposed. The directional distribution of the feathers is perturbed to make the feathers in the coil bending area exhibit a more natural diffusion trend; among which, the curvature From the center line It is obtained by calculating the derivative, specifically... Defined as the magnitude of the derivative of the tangent vector, its calculation principle is shown in equation (5): ； in, That is Next, calculate the second normal vector. In the Frenet frame, By tangent vector and principal normal vector The definition of the cross product, and in the revised Bishop frame In the middle, still satisfied The calculation principle is shown in equation (6): Right now, ; Finally, regarding particle path control, in order to maintain the overall directional constraint while exhibiting flexible fiber characteristics, the final hair generation path, i.e., the initial growth direction of the hair, is obtained. The calculation principle is shown in equation (7): ； in, This is the normal-dominant factor, ensuring that hair grows along the normal direction of the yarn surface. This is the tangential offset coefficient, used to control the degree of inclination of the hairs along the yarn axis. The curvature perturbation coefficient adjusts the diffusion intensity of hairs in the bending region.

6. The geometric simulation and high-realism rendering method for knitted fabrics as described in claim 5, characterized in that: In step 6, firstly, the growth direction of the hair is dominated by the yarn section normal, thereby determining the reference growth axis of each hair in three-dimensional space; then, the tilt angle of the hair is finely adjusted by linear interpolation, thereby simulating the unidirectional or helical tilt characteristics formed by the fiber during twisting and spinning. Finally, fine-tuning is performed based on tangentiality and curvature to obtain the three-dimensional path of a single hair. The calculation principle is shown in equation (8): ； in, It is the length parameter of the feathers themselves, defined in the interval ∈[0, 1], indicating from the root of the hair. =0 to the tip =1 normalized position; u(t, This represents the mapping of parameter t to a specific yarn position and the position of a specific length of hairiness. The displacement vector of the disturbance superimposed at the location.

7. The geometric simulation and high-fidelity rendering method for knitted fabrics as described in claim 1, characterized in that: In step 7, firstly, the density, length, and softness of the feathers are manually adjusted according to the yarn thickness, local stress, and the different contact areas of the coils; then, from the perspective of visual performance and function, the feather system is layered into a three-layer overlapping structure of down layer, feather layer, and free feathers; finally, in the real-time interaction stage, the natural movement and physical interaction of the feathers are simulated through force field control and collision detection.

8. The geometric simulation and high-realism rendering method for knitted fabrics as described in claim 1, characterized in that: The specific implementation method of step 8 is as follows: First, a microscopic optical representation layer is constructed by integrating physical displacement maps and hair bidirectional scattering distribution functions. The grayscale height information of the displacement maps is used to replace the geometry of the yarn substrate, accurately shaping the spiral twisting texture between individual strands and the macroscopic surface unevenness. Simultaneously, a hair bidirectional scattering distribution function shader is configured, using its dual-mirror reflection model to simulate the reflection, refraction, and internal reflection behavior of light on the surface of microfibers. A volume scattering algorithm is introduced to reproduce the unique semi-transparent edges and soft halo effect of knitted fabrics by calculating the mean free path and light energy transfer law inside the hair layer. Subsequently, the multi-layered hair structure was sampled for spatial position and extracted for orientation field, baked into a shell texture with multi-level depth, and a visual substitution of tens of thousands of fibers was achieved by using a limited geometric patch in combination with a transparency mask to obtain a knitted fabric simulation model; at the same time, the knitted fabric simulation model was optimized and compressed by geometric mesh topology. Finally, the optimized knitted fabric simulation model was exported using a standard model exchange format, and the feather material was reconstructed. By using a custom shader to approximate the anisotropic reflection of the knitted fabric simulation model, and combining it with real-time global illumination and shadow algorithms, we can ensure that the knitted fabric simulation model can maintain stable visual fidelity and physical texture under different dynamic perspectives.

9. A geometric simulation and high-fidelity rendering system for knitted fabrics, characterized in that: It includes a processor and a memory, the memory being used to store program instructions, and the processor being used to call the program instructions in the memory to execute the geometric simulation and high-realism rendering method for knitted fabrics as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, It includes a readable storage medium on which a computer program is stored, and when the computer program is executed, it implements a geometric simulation and high-realism rendering method for knitted fabrics as described in any one of claims 1-8.

Images