Interactive generation and arrangement method for three-dimensional martial arts action sequence

Abstract

The invention belongs to the technical field of three-dimensional images, and discloses an interactive generation and arrangement method for a three-dimensional martial arts action sequence. Comprising the following steps: acquiring skeletal animation data of a martial art character, defining a plurality of virtual stiffness transmitters on a martial art character skeletal system corresponding to the skeletal animation data, binding each virtual stiffness transmitter to a preset skeletal node, constructing a stiffness propagation curve along a character skeletal link, and calculating a stiffness flow field vector field on each time sampling point; setting a physical agent model; in a sequence playing process, converting the vector field of the force flow field into direction force and torque applied to each clothes control point in a physical agent model, and performing simplified collision detection and response on main limbs of the martial art role; obtaining a motion trail of each clothes control point; and mapping to a low-resolution clothing preview model to generate a clothing dynamic deformation effect, thereby realizing interactive generation and arrangement for the three-dimensional martial arts action sequence.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional image technology, and more specifically, to an interactive generation and choreography method for three-dimensional martial arts movement sequences. Background Technology

[0002] Tai Chi, as the most widely practiced martial art, has been extensively used in games and films. However, the core characteristic of Tai Chi is internal strength (Nuan Jin). In 3D Tai Chi animation, demonstrating this internal strength is crucial. Therefore, the movement of clothing caused by "relaxation and sinking" (such as the swaying of loose Tai Chi sleeves and trousers) is a key visual signal conveying the flow of "strength" and the character's weight. The movement of clothing is not solely driven by external collisions with the limbs or by wind, but originates from internal "strength." Due to "relaxation and sinking," the clothing's response to limb movements is delayed and gentle, rather than an immediate follow-through like a tight-fitting garment. For example, a single exertion of force (such as a concealed hand strike) causes the sleeves to produce a wave-like decaying motion that lasts much longer than the force exertion itself. In movements like cloud hands, the sleeves should form a distinct spiral or arc trajectory, rather than a simple left-right sway. This force manifests as a continuous "expansion" or "rotation" force field emanating from the core of the torso and spiraling along the limbs. This is something that traditional methods based on vertex skinning or simple physics-based clothing simulations cannot perceive or respond to.

[0003] Based on the above problem description, an interactive generation and choreography method for three-dimensional martial arts movement sequences is designed. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: an interactive generation and choreography method for three-dimensional martial arts movement sequences, comprising: S1. Obtain the skeletal animation data of the martial arts character. The skeletal animation data includes the position, linear velocity, and angular velocity of each bone joint on the time axis. S2. Define multiple virtual force emitters on the skeletal system of the martial arts character corresponding to the skeletal animation data. Each virtual force emitter is bound to a preset bone node and has force intensity parameters and force direction parameters that can be adjusted by the animator. The force intensity parameters are automatically associated with the linear velocity and angular velocity of the corresponding bone. S3. Based on the parameters of the virtual force emitter and the skeletal animation data, construct the force propagation curve along the character's skeletal link, and calculate the force flow field vector field at each time sampling point; S4. Set up a physical proxy model consisting of a small number of clothing control points. Each clothing control point has mass, damping and adjacent point constraint relationships. S5. During the sequence playback, the force flow field vector field of each time step is converted into the directional force and torque applied to each clothing control point in the physical proxy model, and at the same time, simplified collision detection and response are performed on the main limbs of the martial arts character. S6. Perform physical simulation on the physical agent model to obtain the motion trajectory of each clothing control point; S7. Map the motion of the physical proxy model to the low-resolution clothing preview model in real time to generate dynamic clothing deformation effects. S8. Receive the animator's input for adjusting the force parameters in the interactive editing interface, and update the motion of the force flow field vector field, physical proxy model, and clothing preview model accordingly.

[0005] Preferably, the method of defining multiple virtual force emitters on the skeletal system of the martial arts character corresponding to the skeletal animation data includes: Virtual force emitters are bound to the preset skeletal nodes, and the local coordinate system of each virtual force emitter is aligned with the local coordinate system of the corresponding skeletal node. The skeletal nodes include the character's dantian, mid-spine, shoulders, elbows, wrists, hips, knees, and ankles. Each virtual force emitter is defined to include at least a force intensity parameter and a force direction parameter. The force intensity parameter includes explosive force parameter, continuous force parameter and spiral force parameter, and the force direction parameter is related to the character's current orientation. Based on the linear and angular velocities of the corresponding bone nodes in the skeletal animation data, the explosive force and continuous force parameters are automatically initially estimated, and animators are allowed to refine the explosive force, continuous force, and spiral force parameters through time curves.

[0006] Preferably, S3 includes: At each time sampling point, the force intensity parameters and force direction parameters of each virtual force emitter are projected onto the tangential and normal directions of the corresponding skeletal link to obtain the local force vector propagating along the skeletal link; Along the skeletal link from the torso to the extremities, the local force vector is accumulated and smoothed to form a force propagation curve that changes continuously over time. On the geometry surrounding the martial arts character or the geometry surrounding the clothing, surface sampling points are sampled according to a preset density. The force propagation curve is diffused to the surface sampling points through a radial attenuation function and an angle weighting function to obtain the force flow vector field near the outer contour of the character.

[0007] Preferably, the physical proxy model consisting of a small number of clothing control points is set up by the following method: Based on the topology and attachment skeleton of the high-resolution clothing mesh, several vertex positions representing key parts such as cuffs, pant legs, hem, and shoulders are selected as the initial positions of the clothing control points. On the surface of the high-resolution clothing mesh, control point links are defined along the main direction from the cuff to the shoulder, and the initial positions of the clothing control points are expanded to a preset number of clothing control point sets. A physical proxy model is constructed by assigning mass, linear damping, local elastic coefficient, and distance constraint parameters between adjacent clothing control points to each clothing control point.

[0008] Preferably, the method for converting the force flow field vector field at each time step into directional force and torque applied to each clothing control point in the physical proxy model includes: At each time step, based on the position of each clothing control point in the character's local coordinate system, query the force flow vector field at the corresponding location; Based on the decomposition results of the local normal and tangential directions of the clothing control points, the queried force flow field vectors are mapped to tensile force and tangential torque respectively, and gravity and inertial force caused by skeletal animation data are superimposed. Apply a relatively amplified amount of cotton force to the garment control points at the cuffs and trouser legs.

[0009] Preferably, the method for simplified collision detection and response of the main limbs of a martial arts character includes: At the skeletal nodes of the main limbs of a martial arts character, a spherical collider is added. The size of the spherical collider is determined by the local geometric bounding volume of the corresponding skeletal node. At each time step, for each clothing control point, calculate the shortest distance from it to the surface of the nearby spherical collider. When the shortest distance is less than zero, trigger a collision response. In the collision response, the position of the clothing control point is projected onto the outer side of the spherical collider surface along the collision normal direction, and the linear velocity of the clothing control point is decomposed into normal and tangential components. The normal component is reflected and attenuated, and the tangential component is multiplied by the friction coefficient.

[0010] Preferably, the method for reflecting and attenuating the normal component includes: Calculate the incident velocity vector of the clothing control point relative to the surface of the sphere colliding body; decompose the incident velocity vector into a normal component perpendicular to the collision surface and a tangential component parallel to the collision surface; For the normal component, reverse its direction and multiply it by a preset bounce coefficient to obtain a new normal component; The new normal component is added to the tangential component processed by the friction coefficient, and then recombined to obtain the new velocity of the clothing control point after the collision response.

[0011] Preferably, S7 includes: For each vertex position of the low-resolution clothing model, pre-calculate its interpolation weights to the nearest few clothing control points, and express the vertex position displacement as a weighted sum of the displacements of the corresponding clothing control points; During runtime, at each time step, the vertex positions of the low-resolution clothing model are updated based on the real-time positions of the clothing control points and the pre-calculated interpolation weights.

[0012] Preferably, the method for updating the vertex positions of the low-resolution clothing model based on the real-time positions of the clothing control points and the pre-calculated interpolation weights includes: During the initialization phase, interpolation weights are bound to each vertex position of the clothing preview model; At each simulation time step, obtain the current global coordinates of all clothing control points; For each vertex position of the clothing preview model, based on its interpolation weight, calculate the weighted average displacement of the current clothing control point for each associated clothing control point; The weighted average displacement value is superimposed on the initial position coordinates of the vertex to obtain the final position of the vertex in the current frame; Iterate through all the vertex positions of the clothing preview model, synchronize the update of its position coordinates, and generate the dynamic deformation effect of the clothing.

[0013] Preferably, the method for receiving the animator's input on adjusting the force parameters in the interactive editing interface includes: In the timeline editor, the explosive force parameter curve, the continuous force parameter curve, and the spiral force parameter curve of each virtual force launcher are displayed in the form of curves, allowing animators to modify the curve shape by editing control points; In the 3D view, the position and current force direction parameters of the virtual force emitter are displayed in the form of interactive arrows, allowing animators to directly change the initial estimates of the force direction and force intensity parameters in 3D space. After the animators adjusted the force intensity parameters, force direction parameters, and force time curve, the simulation results of the force flow field and physical proxy model were recalculated only for the affected time interval and the relevant skeletal links.

[0014] The technical effects and advantages of this invention's interactive generation and choreography method for three-dimensional martial arts movement sequences are as follows: By transforming the core and most abstract concept of "internal strength" in Tai Chi into a complete technical pipeline encompassing data definition, spatial calculation, physical response, and visual presentation, the effect of "internal strength driving clothing" has been turned into a standardized industrial technology. By employing a physical proxy model and a local recalculation strategy, interactive editing and real-time previewing of complex effects become possible while ensuring that clothing dynamics conform to physical laws and martial arts characteristics (long, spiraling motions). This addresses a core pain point in pre-production previews and rapid prototyping for film and games. It significantly reduces the technical barriers and time costs associated with producing high-quality martial arts animation, especially animations depicting the essence of internal martial arts. Animators no longer need to be martial arts experts to create movements that conform to the principles of martial arts by adjusting parameters.

[0015] By using parameterization and templates, it is possible to ensure that the animation styles produced by different production staff within the same project are consistent, and to establish an internal "power animation library" for asset reuse.

[0016] The output standard skeletal animation data and optional proxy motion trajectories can be seamlessly integrated with downstream game engines or render farms.

[0017] The solution is interconnected, forming a complete and self-consistent technical system from force parameters to proxy model construction and collision response. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the interactive generation and choreography method for three-dimensional martial arts movement sequences according to the present invention; Figure 2 A schematic diagram of a simplified collision detection and response method for the main limbs of a martial arts character in this invention; Figure 3 This is a schematic diagram illustrating the method of receiving animator's input for adjusting force parameters in an interactive editing interface according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 , Figure 2 and Figure 3 In this embodiment of the invention, an interactive generation and choreography method for three-dimensional martial arts movement sequences includes: S1. Obtain the skeletal animation data of the martial arts character. The skeletal animation data includes the position, linear velocity, and angular velocity of each bone joint on the time axis in the three-dimensional martial arts action sequence. S2. Define multiple virtual force emitters on the martial arts character bone system corresponding to the skeletal animation data. Each virtual force emitter is bound to a preset bone node and has an animator-adjustable force intensity parameter and force direction parameter. The force intensity parameter is automatically associated with the linear velocity and angular velocity of the corresponding bone; S3. Based on the parameters of the virtual force emitters and the skeletal animation data, construct a force propagation curve along the character bone link and calculate the force flow vector field at each time sampling point; S4. Set up a physical proxy model composed of a small number of clothing control points. Each clothing control point has mass, damping, and adjacent point constraint relationships; S5. During the sequence playback, convert the force flow vector field at each time step into the directional forces and torques applied to each clothing control point in the physical proxy model, and at the same time perform simplified collision detection and response on the main limbs of the martial arts character; S6. Perform physical simulation on the physical proxy model to obtain the motion trajectories of each clothing control point; S7. Map the motion of the physical proxy model to the low-resolution clothing preview model in real time to generate the clothing dynamic deformation effect; S8. Receive the animator's adjustment input of the force parameters in the interactive editing interface, and accordingly update the motion of the force flow vector field, the physical proxy model, and the clothing preview model, realizing the interactive choreography and real-time preview of the action sequence.

[0021] The force of Tai Chi is abstract and internal. Traditional animation production relies on the animator's subjective understanding of key frames and cannot systematically define, adjust, and reuse the force. It is difficult to accurately distinguish and quickly produce the force characteristics of different schools and different moves. That is the core contradiction of "it is difficult to embody the internal force" in the background technology. In this design, traditional martial arts concepts such as explosion, softness, and spiral are clearly defined as scalar parameters that can be independently adjusted, making the internal force quantifiable. Binding the emitter to the key points of mechanical conduction such as the dantian and joints makes the starting point of the internal force conform to human motion science and provides a physical basis for subsequent force propagation calculations. Automatically estimating the initial parameters using bone velocity greatly reduces the entry threshold for animators; at the same time, providing a time curve for fine-grained editing achieves the balance between efficiency and artistic control. This enables different styles to be quickly generated through parameter differences. The specific design is as follows: The method of defining multiple virtual force emitters on the martial arts character bone system corresponding to the skeletal animation data includes: Virtual force emitters are bound to preset skeletal nodes. The local coordinate system of each virtual force emitter is aligned with the local coordinate system of the corresponding skeletal node (automatically defined by the 3D animation software based on the bone orientation and preset axis, usually with the bone length direction as one principal axis (such as the Y-axis), and the lateral and normal directions of the joint as the other two axes). The skeletal nodes are selected based on the principles of martial arts biomechanics, including the character's dantian (the skeletal node near the center of gravity), the middle of the spine (such as the thoracic or lumbar vertebrae), both shoulders, both elbows, both wrists, both hips, both knees, and both ankles. The binding operation refers to setting the spatial transformation matrix of each virtual force emitter and the transformation matrix of the corresponding skeletal node to a fixed parent-child hierarchical relationship. The alignment operation is specifically: the origin of the coordinates of the virtual force emitter is set at the location of the skeletal node, and the directions of its coordinate axes are consistent with the directions of the local coordinate axes of the skeletal node, thereby ensuring that the virtual force emitter can move and rotate synchronously with the movement of the bones. Each virtual force emitter is defined to include at least a force intensity parameter and a force direction parameter. The force intensity parameter is a scalar or vector parameter used to quantify the strength of the force output, including an explosive force parameter (representing a short and high-intensity impact force, mainly affecting the peak force field at the moment of action initiation), a continuous force parameter (representing a sustained and gentle force, mainly affecting the decay time and subsequent fluctuations of the force field), and a spiral force parameter (representing a force with rotational characteristics, affecting the rotational component of the force field in the form of torque). The force direction parameter is related to the character's current orientation; the force direction parameter is a vector or Euler angle parameter used to define the direction of force emission in the local space of the virtual force emitter; its default direction is related to the local axis of the skeletal node (such as the forward direction), and can be dynamically calculated in combination with the character's overall world space orientation or the target direction specified by the animator, so that the force direction can change naturally with the character's turning movements, etc. Based on the linear and angular velocities of corresponding bone nodes in the skeletal animation data, the explosive force and continuous force parameters are automatically initially estimated. Animators can then refine these parameters using time curves to define different types of force characteristics in various martial arts movements (e.g., concealed hand and elbow strike, cloud hands). Specifically: the magnitude of the linear velocity vector (i.e., rate of change of position) and the magnitude of the angular velocity vector (i.e., rate of change of rotation) of the bone nodes at each time sampling point are calculated; the linear velocity magnitude is smoothed and normalized, then mapped to the initial estimated value of the explosive force parameter to reflect the intensity of the movement; the angular velocity magnitude is mixed with the linear velocity magnitude according to preset weights, and after low-pass filtering, mapped to the initial estimated value of the continuous force parameter to reflect the rotational and sustained characteristics of the movement; the estimated parameter values ​​are presented in the form of time curves. The interactive editing interface provides animators with a time curve editor for explosive force, soft force, and spiral force parameters. The time curve editor displays the curves of each parameter changing over time on the timeline. Animators can directly modify the curve shape by adding, deleting, or dragging keyframes on the curve, thereby finely editing the rhythm, amplitude, and waveform of the force intensity changing over time. By adjusting these three curves separately, animators can create a force characteristic dominated by explosiveness and rapid decay in movements such as "Covering Hand Elbow Strike," or a force characteristic dominated by soft and spiral force and slow circulation in movements such as "Cloud Hands," thus defining diverse force expressions in different martial arts action segments.

[0022] The force doesn't act on the bones, but rather passes through the flesh and affects the outer clothing. Therefore, a method is needed to transform the point-like force parameters on discrete bone nodes into a continuous, physically intuitive (e.g., conducted along bones, attenuating with distance) field filling the space surrounding the character. This design addresses the issue of force manifesting as a force field emanating from the core of the torso and spiraling along the limbs. By projecting and accumulating along the tangential / normal axes of the bone links, it simulates the force transmission path of internal martial arts. Hysteresis and softness are introduced through smoothing filtering; radial attenuation and angular weighting functions simulate the diffusion and weakening of force in space, avoiding abrupt changes in the force field. The generated force flow vector field provides a unified query interface for subsequent force calculations at each clothing control point. The specific design is as follows: S3 includes: At each sampling point in time, i.e., at a fixed time interval inherent to the skeletal animation data or set by the user, such as 30 frames per second or 60 frames per second, the force intensity parameters and force direction parameters of all virtual force emitters at the current moment are obtained. The force intensity parameters and force direction parameters of each virtual force emitter are projected onto the tangential and normal directions of the corresponding skeletal link to obtain the local force vector propagating along the skeletal link. Specifically, for each virtual force emitter, firstly, in its local coordinate system, an original force vector is calculated based on its force intensity parameters and force direction parameters; then, the skeletal link to which the skeletal node bound to the emitter belongs is determined. It is a linear chain composed of a series of skeletal nodes connected by parent-child relationships. The tangential direction (usually the direction along the bone pointing to the next child bone or parent bone) and normal direction (perpendicular to the tangential direction and located in the plane defined by the local coordinate system of the bone, usually representing the possible direction of muscle expansion or lateral force) of the skeletal node on its respective skeletal chain are calculated. Finally, the original force vector is decomposed and projected onto these two directions to obtain the tangential component that is transmitted along the bone and the normal component that is diffused perpendicular to the bone, respectively. The two together constitute the local force vector propagating along the skeletal chain, which is supplemented with direction and magnitude information.

[0023] Along the skeletal pathway from the torso to the extremities, local force vectors are accumulated and smoothed to form a force propagation curve that changes continuously over time. Specifically, starting from the root skeletal pathway (e.g., the spine), the process proceeds link by link towards the extremities (e.g., fingers, toes). For each skeletal pathway, within a given time sampling point, the tangential components of all local force vectors generated by virtual force emitters are vector-superimposed, and the normal components are averaged to obtain the synthetic local force vector for that pathway at that moment. Then, in the time dimension, at multiple consecutive time sampling points, a smoothing filtering algorithm is applied to the synthetic local force vector sequence for that pathway. For example, a moving average filter is used to remove high-frequency noise caused by data sampling or minor movement fluctuations, simulating the smoothness and sluggishness of internal force transmission. The filtered and chronologically ordered synthetic local force vector sequence then constitutes a force propagation curve that changes continuously over time, used to describe the force propagation state and intensity along that skeletal pathway.

[0024] On the geometry surrounding the martial arts character or the geometry surrounding clothing, sampling points on the surface are sampled according to a preset density (referring to a series of three-dimensional spatial points uniformly distributed on the surface of the surrounding geometry or on a shell offset from the surface according to the required flow field accuracy). The force propagation curve is diffused to the surface sampling points through a radial attenuation function and an angle weighting function, resulting in a force flow field vector field near the character's outer contour. The geometry surrounding the martial arts character refers to a simplified three-dimensional geometry that tightly wraps around the entire character (including the body and unclothed parts), such as a collision-bounded body composed of multiple capsules or convex hulls. The geometry surrounding clothing refers to a simplified three-dimensional geometry that specifically wraps around the area of ​​clothing worn by the character, and its axially aligned bounding box or convex hull can be calculated based on the initial pose of the high-resolution clothing mesh. The specific implementation steps are as follows: For each surface sampling point, traverse the force propagation curves of all skeletal links; for the value of each curve at the current sampling point (a spatial vector), calculate the spatial distance from the surface sampling point to the nearest point on the skeletal link, and the angle of the surface sampling point relative to the tangential direction of the bone at the nearest point; apply a radial decay function, whose value monotonically decreases with increasing distance (e.g., using exponential decay), to simulate the effect of force weakening with distance; apply an angle weighting function, which adjusts the influence weight according to the angle (e.g., the diffusion weight perpendicular to the skeletal tangential is less than the weight along the tangential direction), to simulate the directionality of force propagation; multiply the local force vector of the current link by the weights calculated by these two functions to obtain the contribution vector of the link to the surface sampling point; sum the contribution vectors of all skeletal links to finally obtain the composite vector of the surface sampling point at the current moment; The composite vector set of all surface sampling points at a certain moment is used to define a discrete vector field covering the character's outer space. This force flow vector field intuitively represents the magnitude and direction distribution of the "internal force" generated by the internal skeletal movement in the space around the character, which is used to drive the subsequent clothing physics simulation.

[0025] Physical simulations of high-precision clothing meshes (such as using the finite element method) are computationally expensive, making them unsuitable for interactive editing. However, simple skinning or rigid body simulations cannot represent the soft inertia, hysteresis, and undulation of fabric. This design selects control points through high-poly topology and key areas (cuffs, hems) to ensure the surrogate model captures large deformations and key dynamic characteristics of the clothing. Tens of thousands or even hundreds of thousands of mesh vertices are simplified into a mass-spring system composed of dozens to hundreds of control points, enabling real-time physical simulation. Physical properties such as mass, damping, and elasticity are assigned to the control points, making their dynamic behavior macroscopically approximate that of real fabric, laying the physical foundation for subsequent force field actuation. The specific design is as follows: The physical proxy model, consisting of a small number of clothing control points, is configured using the following methods: Based on the topology and attachment skeleton of the high-resolution clothing mesh, several vertex positions representing key parts such as cuffs, trouser legs, hems, and shoulders are selected as the initial positions of clothing control points. Specifically, the topology of the high-resolution clothing mesh is analyzed to identify its boundary loops and high-curvature regions to locate clothing openings or free ends such as cuffs, trouser legs, and hems. Simultaneously, the bone weight data of the clothing mesh vertices is read. Bone weight refers to the weight of each vertex affected by the movement of one or more bone nodes. Based on the bone weights and spatial positions, a set of vertices in the shoulder region that are significantly affected by shoulder bone nodes (e.g., those with the highest weight) and are evenly distributed are selected, as well as representative vertices at the boundaries of cuffs, trouser legs, and hems (e.g., vertices evenly spaced on boundary loops). The three-dimensional coordinates of the selected vertices are determined as the initial positions of the clothing control points, enabling them to effectively represent the shape and movement basis of key parts of the clothing.

[0026] On the surface of a high-resolution clothing mesh, control point links are defined along the main direction from the cuff to the shoulder, expanding the initial positions of the clothing control points to a preset number of clothing control point sets. Specifically, a parametric direction field is established on the high-resolution clothing mesh from the cuff to the shoulder. Starting from the selected initial clothing control point at the cuff, sampling is performed along the path indicated by the direction field on the mesh surface (usually along the arm direction, following the fabric texture or seam direction) with a certain sampling step size (which can be dynamically calculated based on the mesh resolution). A series of new intermediate clothing control points are generated by interpolation along the path. This process connects and expands the original discrete keypoints into control point links continuously distributed along the main direction of the fabric (such as the sleeve axis). Similar operations are performed on other parts such as the trouser legs and hem. Finally, all generated clothing control points (including initial and expanded points) are organized according to the link connection relationship to form a set of clothing control points covering the main dynamic area of ​​the clothing, with a preset number of points, far fewer than the number of vertices in the original high-resolution mesh.

[0027] Each clothing control point is assigned a mass (a scalar parameter estimated and allocated based on the local fabric area represented by the control point, used for inertia calculations in the physical simulation), linear damping (a scalar parameter simulating dissipation effects such as air resistance experienced by the fabric during movement, used to attenuate the velocity of the clothing control point), local elastic coefficient (a scalar parameter defining the stiffness of the connection between the clothing control point and its adjacent clothing control points, affecting the fabric's ability to resist tensile or shear deformation), and distance constraint parameters between adjacent clothing control points (a set of constraints defined based on the distance between adjacent clothing control points in the initial posture, used to maintain the basic shape of the fabric in the physical simulation and prevent excessive stretching or compression; this parameter typically includes rest length and constraint stiffness), forming a physical surrogate model.

[0028] How to make the simplified proxy model not only affected by gravity and collision, but also able to "feel" the internal force? This is a problem that needs to be solved. Through spatial interpolation query, each control point can obtain a personalized force input that matches its position. Normal or tangential decomposition mapping transforms the abstract force into concrete force and torque, and superimposes gravity and inertia, so that the movement of clothing is affected by both internal and external forces. The end amplification of the soft force accurately simulates the typical visual characteristics of Tai Chi, such as the continuous decaying swing of the sleeve. A spherical collision body is used to approximate the limbs, and efficient penetration detection and analytical solution (projection method) are performed. The damped sliding of the fabric and the body is simulated by the rebound coefficient and friction coefficient, which avoids serious visual errors while ensuring real-time performance and ensuring the validity of the preview results. The specific design is as follows: The method for converting the force flow field vector field at each time step into directional forces and torques applied to each clothing control point of the physical proxy model includes: At each time step, based on the position of each clothing control point in the character's local coordinate system, the force flow vector field at that position is queried. Specifically, each clothing control point has its own defined three-dimensional spatial coordinates during the simulation. The force flow vector field is defined in space by discrete surface sampling points and their attached vectors. During the query, firstly, in the character's local coordinate system, several (e.g., 3 or more) surface sampling points closest to the clothing control point are found. Then, based on the spatial positions of these neighboring surface sampling points and their recorded force flow vectors, a spatial interpolation algorithm (such as barycentric coordinate interpolation) is used to calculate the estimated value of the force flow vector at the current position of the clothing control point, thus completing the vector query from discrete field to continuous spatial point.

[0029] Based on the decomposition results of the local normal and tangential directions of the clothing control points, the queried force flow field vectors are mapped to tensile force and tangential torque, respectively, and gravity and inertial force caused by skeletal animation data are superimposed. Specifically, firstly, a local coordinate system is established for each clothing control point, and the average value of the vertex normals of the high-resolution clothing mesh near the point is taken as the normal direction, representing the vertical direction of the fabric surface; in the plane perpendicular to this normal, the velocity direction along the link of the clothing control point is selected as the tangential direction.

[0030] Then, the retrieved force flow field vector is decomposed into the local coordinate system: the vector is projected onto the normal direction to obtain the normal component; the vector is projected onto the tangential direction to obtain the tangential component.

[0031] Among them, mapping to tensile force means: directly taking the normal component value as the magnitude of the force, with the direction along the normal direction, and acting on the control point of the garment.

[0032] Mapping to tangential torque means converting the magnitude of the tangential component into the magnitude of the torque, with its rotation axis being the normal direction of the point, and the rotation direction being determined by the direction of the tangential component. This torque is applied to the clothing control point.

[0033] The superposition of inertial force caused by skeletal animation data refers to: calculating the inertial force generated by the non-uniform motion of the clothing control point due to following the skeleton based on the acceleration bound to the clothing control point, and superimposing it as an external force onto the point.

[0034] Apply a relatively amplified (meaning enhanced according to preset rules or proportions) force component to the clothing control points at the cuffs and trouser legs to maintain the continuously decaying swing trajectory of the sleeves and trouser legs after the explosive movement ends.

[0035] The method for simplified collision detection and response of the main limbs of martial arts characters includes: At the skeletal nodes of the main limbs of a martial arts character, a spherical collider is added. The size of the spherical collider (to ensure that the spherical collider can cover the volume of the limb corresponding to the bone) is determined by the local geometric bounding volume of the corresponding bone node. Specifically, the spherical collider is a sphere used to approximate the shape of the character's limbs for rapid intersection testing.

[0036] The additional method is as follows: for the main limbs that need to be collision detected, create a spherical geometry object for each of the corresponding key bone nodes, and fix the center of the spherical object at the current position of the bone node, making it a spatial appendage of the bone.

[0037] At each time step, for each clothing control point, calculate the shortest distance from it to the surface of a nearby spherical collider. When the shortest distance is less than zero, trigger a collision response. For each detected spherical collider, the shortest distance is calculated as (vector length from the clothing control point to the center of the sphere) - (radius of the spherical collider).

[0038] In the collision response, the position of the clothing control point is projected onto the outer surface of the spherical collider along the collision normal direction. (First, the collision normal direction is calculated; this direction is a unit vector pointing from the center of the spherical collider to the current position of the clothing control point. Then, the position of the current clothing control point is moved along this collision normal direction until it is exactly on the surface of the spherical collider, i.e., the new position satisfies: new position = center position + collision normal direction × radius of the spherical collider, thus completing the projection.) The linear velocity of the clothing control point is decomposed into normal and tangential components. The normal component is subjected to reflection and attenuation processing, and the tangential component is multiplied by the friction coefficient.

[0039] The method for reflecting and attenuating the normal component includes: Calculate the incident velocity vector of the clothing control point relative to the surface of the spherical collider; specifically, upon triggering the collision response, record the initial velocity vector of the clothing control point and define this velocity vector as the incident velocity vector. This calculation is an absolute value calculation relative to the character's local coordinate system, but the physical meaning of this vector is understood as the approach velocity of the clothing control point relative to the stationary spherical collider surface.

[0040] The incident velocity vector is decomposed into a normal component perpendicular to the collision surface and a tangential component parallel to the collision surface. For the normal component, reverse its direction (this is a mathematical vector operation; simply multiply it by -1), and multiply it by a preset bounce coefficient to obtain a new normal component, where the bounce coefficient is a value between 0 and 1, used to simulate the damping and energy absorption of clothing collisions; The new normal component is added to the tangential component processed by the friction coefficient, and then recombined to obtain the new velocity of the clothing control point after the collision response.

[0041] The physical proxy model has only a few control points, and its motion trajectory is discrete. How to smoothly propagate the motion of these points to the preview model, which has a much higher vertex count, is a problem that needs to be solved. This design pre-calculates the interpolation weights from each preview vertex to the nearest control point. At runtime, only a small amount of weighted summation calculation is needed to drive the deformation of the entire high-vertex mesh, resulting in extremely high efficiency. The multi-control-point weighted averaging mechanism ensures smooth and continuous mesh deformation, avoiding local collapse or sharp distortions that may occur due to too few control points, thus guaranteeing the credibility of the preview visual effect. The lightweight mapping workflow allows animators to see the updated clothing deformation effect in a very short time after adjusting parameters, forming a rapid iterative closed loop of adjustment-preview. The specific design is as follows: S7 includes: For each vertex of the low-resolution clothing model, pre-calculate its interpolation weights to the nearest few clothing control points, and represent the vertex displacement as a weighted sum of the displacements of the corresponding clothing control points. Specifically, spatial relationship calculations are performed when both the physical proxy model and the clothing preview model are in their initial tethered pose (relaxed pose). For each vertex of the clothing preview model, perform the following operations: In three-dimensional space, calculate the distance from the vertex to the initial position of all clothing control points, and select the K clothing control points with the shortest distance (K is usually 3 or 4) as the set of associated control points that affect the vertex.

[0042] Based on the distance between the vertex and each clothing control point in the associated control point set, the initial influence weight is calculated using the inverse distance weighting method. The formula for calculating the weight w_i is w_i=1÷(d_i^p+ε), where d_i is the distance from the vertex to the i-th associated clothing control point, p is the decay exponent (usually 2), and ε is a very small constant to prevent division by zero.

[0043] The calculated K initial weights are normalized to ensure that the sum of all weights is 1, thus obtaining the interpolation weights.

[0044] Store a list of indices of its associated control point set and a list of corresponding interpolation weights for each vertex. Vertex displacement = ∑(W_i × displacement of the i-th associated control point).

[0045] During runtime, at each time step, the vertex positions of the low-resolution clothing model are updated based on the real-time positions of the clothing control points and the pre-calculated interpolation weights.

[0046] The method for updating the vertex positions of the low-resolution clothing model based on the real-time positions of the control points and the pre-calculated interpolation weights includes: During the initialization phase, interpolation weights are bound to each vertex of the clothing preview model; At each simulation time step, obtain the current global coordinates of all clothing control points; For each vertex in the clothing preview model, based on its interpolation weights, the weighted average displacement of the current clothing control point is calculated for each associated clothing control point. Specifically, the displacement vector of each associated clothing control point is multiplied by its corresponding interpolation weight, and then all these weighted displacement vectors are summed. The summation result is the weighted average displacement value experienced by that vertex.

[0047] The weighted average displacement value is superimposed on the initial position coordinates of the vertex to obtain the final position of the vertex in the current frame; the sum is the final position that the vertex should appear in the preview screen after experiencing all the physics simulations and force drives in the current frame.

[0048] Iterate through all vertices of the clothing preview model, synchronize the update of its position coordinates, and generate dynamic deformation effects of the clothing to achieve real-time preview of clothing deformation.

[0049] The method for receiving animator's input on force parameters in the interactive editing interface includes: The timeline editor displays the explosive force, continuous force, and spiral force parameter curves for each virtual force emitter as curves, allowing animators to modify the curve shapes by editing control points. The timeline editor is a visual component within the interactive editing interface; its horizontal axis represents animation time, and its vertical axis represents parameter values. Within this editor, a timeline curve is drawn for each force intensity parameter component (explosive force, continuous force, spiral force) for each virtual force emitter. This curve is formed by connecting a series of curve control points defined on the time-value plane. Animators can perform direct operations through the graphical user interface, including but not limited to: dragging existing curve control points to change their corresponding time point parameter values; adding new curve control points to introduce more complex parameter changes; deleting existing curve control points to simplify the curve; and adjusting the tangent handles of curve control points to change the smoothness and direction of the curve at that point. These operations update the underlying force intensity parameter numerical sequence in real time, allowing for intuitive and precise control over the pattern of force changes over time.

[0050] In the 3D view, the position and current force direction parameters of the virtual force emitter are displayed in the form of interactive arrows, allowing animators to directly rotate the arrows or adjust their length in 3D space to change the initial estimates of the force direction and force intensity parameters. After the animators adjusted the force intensity parameters, force direction parameters, and force time curve, the simulation results of the force flow field and physical proxy model were recalculated only for the affected time interval and the relevant skeletal links.

[0051] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0052] It should be noted that all formulas in this manual are calculated by removing dimensions and taking their numerical values. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0053] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for interactive generation and choreography of a sequence of three- dimensional martial arts movements, characterized in that, The method comprises the following steps: S1, acquiring martial arts character bone animation data, the bone animation data comprising positions, linear velocities and angular velocities of each bone joint on a time axis; S2, defining a plurality of virtual force emitters on a martial arts character bone system corresponding to the bone animation data, each virtual force emitter being bound to a preset bone node and having an intensity parameter and a direction parameter of force, which are adjustable by an animator, the intensity parameter being automatically associated with the linear velocity and the angular velocity of the corresponding bone; S3, constructing a force propagation curve along a character bone link and calculating a force flow field vector field at each time sampling point based on the parameters of the virtual force emitters and the bone animation data; S4, setting a physical agent model composed of a small number of clothing control points, each clothing control point having a mass, a damping and a neighboring point constraint relationship; S5, in a sequence playback process, converting the force flow field vector field at each time step into a directional force and a torque applied to each clothing control point in the physical agent model, and simultaneously performing simplified collision detection and response on the main limbs of the martial arts character; S6, performing physical simulation on the physical agent model to obtain the motion trajectory of each clothing control point; S7, mapping the motion of the physical agent model to a low-resolution clothing preview model in real time to generate a dynamic deformation effect of the clothing; S8, receiving an adjustment input of the force parameters by the animator in an interactive editing interface, and updating the motion of the force flow field vector field, the physical agent model and the clothing preview model accordingly.

2. The method for interactive generation and choreography of three-dimensional martial arts sequences according to claim 1, wherein, The method of defining a plurality of virtual force emitters on a martial arts character bone system corresponding to the bone animation data comprises: binding a virtual force emitter at each preset bone node, aligning the local coordinate system of each virtual force emitter with the local coordinate system of the corresponding bone node, wherein the bone nodes include the character's dantian, the middle of the spine, the shoulders, the elbows, the wrists, the hips, the knees and the ankles; defining each virtual force emitter to include at least an intensity parameter and a direction parameter of force, wherein the intensity parameter includes an explosive force parameter, a soft force parameter and a spiral force parameter, and the direction parameter is related to the current orientation of the character; automatically initially estimating the explosive force parameter and the soft force parameter according to the linear velocity and the angular velocity of the corresponding bone node in the bone animation data, and allowing the animator to refine and edit the explosive force parameter, the soft force parameter and the spiral force parameter through a time curve.

3. The interactive generation and choreography method for three-dimensional martial arts sequences according to claim 2, characterized in that, The S3 comprises: projecting the intensity parameter and the direction parameter of force of each virtual force emitter to the tangent and normal directions of the corresponding bone link at each time sampling point to obtain a local force vector propagating along the bone link; accumulating and smoothing the local force vector in the direction of the bone link from the torso to the end limb to form a force propagation curve that changes continuously with time; diffusing the force propagation curve to the surface sampling points on the martial arts character bounding geometry or the clothing bounding geometry according to a preset density to obtain a force flow field vector field near the outer contour of the character by a radial decay function and an angle weight function.

4. The interactive generation and choreography method for three-dimensional martial arts sequences according to claim 3, characterized in that, The method of setting a physical agent model composed of a small number of clothing control points comprises: According to the topology of the high-resolution clothing mesh and the hanging skeleton, a plurality of vertex positions representing key positions of sleeves, legs, hems and shoulders are selected as initial positions of clothing control points; On the surface of the high-resolution clothing mesh, a control point link is defined along a main direction from the sleeve to the shoulder, and the initial positions of the clothing control points are expanded to a preset number of clothing control point sets; Each clothing control point is assigned a mass, a linear damping, a local elastic coefficient and a distance constraint parameter between adjacent clothing control points to form a physical agent model.

5. The interactive generation and choreography method for three-dimensional martial arts sequences according to claim 4, characterized in that, The method for converting the force flow field vector field of each time step into a directional force and a torque applied to each clothing control point in the physical agent model comprises: At each time step, each clothing control point queries the force flow field vector corresponding to its position in the local coordinate system of the character according to its position in the local coordinate system of the character; According to the decomposition results of the local normal direction and the tangent direction of the clothing control point, the queried force flow field vector is respectively mapped into a stretching directional force and a tangential torque, and the gravity and the inertial force caused by the skeletal animation data are superimposed; A relatively amplified soft force component is applied to the clothing control points at the ends of the sleeves and the legs.

6. The interactive generation and choreography method for three-dimensional martial arts sequences according to claim 5, characterized in that, The method for performing simplified collision detection and response on the main limbs of the martial arts character comprises: A spherical collision body is attached to each skeletal node of the main limbs of the martial arts character, and the size of the spherical collision body is determined by the local geometric bounding volume of the corresponding skeletal node; At each time step, the shortest distance of each clothing control point to the surface of the nearby spherical collision body is calculated, and a collision response is triggered when the shortest distance is less than zero; In the collision response, the position of the clothing control point is projected to the outside of the surface of the spherical collision body along the collision normal direction, the linear velocity of the clothing control point is decomposed into a normal component and a tangent component, the normal component is reflected and attenuated, and the tangent component is multiplied by a friction coefficient.

7. The interactive generation and choreography method for three-dimensional martial arts sequences according to claim 6, characterized in that, The method for performing reflection and attenuation processing on the normal component comprises: The incident velocity vector of the clothing control point relative to the surface of the spherical collision body is calculated, and the incident velocity vector is decomposed into a normal component perpendicular to the collision surface and a tangent component parallel to the collision surface; The direction of the normal component is reversed, and a new normal component is obtained by multiplying the reversed normal component by a preset rebound coefficient; The new normal component and the tangent component processed by the friction coefficient are added to recombine a new velocity of the clothing control point after the collision response.

8. The interactive generation and choreography method for three-dimensional martial arts sequences according to claim 7, characterized in that, The S7 comprises: For each vertex position of the low-resolution clothing model, an interpolation weight to the nearest several clothing control points is pre-calculated, and the displacement of the vertex position is represented as a weighted sum of the displacements of the corresponding clothing control points; At runtime, at each time step, the vertex positions of the low-resolution clothing model are updated according to the real-time positions of the clothing control points and the pre-calculated interpolation weights.

9. The interactive generation and choreography method for three-dimensional martial arts sequences according to claim 8, characterized in that, The method for updating the vertex positions of the low-resolution clothing model according to the real-time positions of the clothing control points and the pre-calculated interpolation weights comprises: In the initialization stage, an interpolation weight is bound to each vertex position of the clothing preview model; At each simulation time step, the current global coordinates of all clothing control points are obtained; For each vertex position of the clothes preview model, according to its interpolation weight, a weighted average value of the current clothes control point displacement is calculated for each clothes control point associated therewith; The weighted average displacement value is superimposed on the initial binding position coordinates of the vertex to obtain the final position of the vertex in the current frame; All vertex positions of the clothes preview model are traversed to complete the synchronous update of their position coordinates, thereby generating a clothes dynamic deformation effect.

10. The interactive generation and choreography method for three-dimensional martial arts sequences according to claim 9, wherein, The method for receiving the adjustment input of the strength parameter by the animator in the interactive editing interface comprises: In the timeline editor, the burst force parameter curve, the soft strength parameter curve and the spiral strength parameter curve of each virtual strength emitter are displayed in the form of curves, and the animator is allowed to modify the curve shape by editing the control points; In the three-dimensional view, the position and the current strength direction parameter of the virtual strength emitter are displayed in the form of interactive arrows, and the animator is allowed to directly change the initial estimates of the strength direction parameter and the strength intensity parameter in the three-dimensional space; After the animator adjusts the strength intensity parameter, the strength direction parameter and the strength time curve, the simulation results of the strength flow field and the physical agent model are recalculated only for the affected time intervals and the related bone links.

Images