Character Model Skinning via Graph Convolution and Nonlinear Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional skinning methods for character models are not applicable to complex character models, as they assume the skeleton closest to a region affects the motion, which is not accurate for models with complex costumes, limiting their application range.
Innovation Solution
A method involving nonlinear transformation and graph convolution processes to determine vertex attributes and connecting relationships, extracting geometric features, and calculating influence probabilities to accurately bind the character model to the skeleton hierarchy, allowing for more complex character models to be skinned effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional skinning method based on distance between grid vertices and joint points is used, then the method is simple to implement, but it cannot be applied to complex character models with costumes
Solution Approach 1:
The patent introduces geometric features (normal vectors, curvatures) and geometric feature operators as intermediary elements between the skeleton hierarchy and the character model vertices. These intermediaries capture the local geometric context of each vertex, enabling the skinning algorithm to make more informed decisions about which skeleton joints should influence each vertex, thereby extending applicability to complex models with costumes while maintaining computational feasibility.
Solution Approach 2:
The patent transforms the traditional distance-based single parameter approach into a multi-parameter system that includes geometric features (normal vectors, curvatures) and their operators. By changing from a simple distance metric to a composite parameter system that incorporates local geometric properties, the method can distinguish between different types of vertices (e.g., those on costume surfaces versus those on character bodies), enabling accurate skinning of complex models.
2Productivity
If manual binding of skeletal joint points to grid vertices is performed, then the binding accuracy can be controlled, but the workload is heavy and time-consuming
Solution Approach 1:
The patent enables the skinning system to automatically determine optimal binding relationships by computing geometric features and their operators for each vertex, then using these features to identify the most relevant skeleton joints. This self-service approach eliminates the need for manual artist intervention in the binding process, dramatically improving productivity while maintaining accuracy through the mathematical rigor of the geometric feature analysis.
Solution Approach 2:
The patent replaces manual artistic judgment with automated parameter-based decision making. By computing geometric features (normal vectors, curvatures) and their operators, the system objectively determines vertex-skeleton relationships based on local geometric properties rather than subjective manual assessment, thereby maintaining binding accuracy while eliminating manual workload.
3Measurement precision
If geometric features and their operators are extracted for each vertex, then the vertex representation becomes more accurate, but the computational complexity increases
Solution Approach 1:
The patent segments the computational task by processing each vertex independently to compute its geometric features and operators. This segmentation allows for parallel computation and optimization, where each vertex's feature extraction can be performed separately and then aggregated, reducing the overall computational burden while maintaining high accuracy in the vertex representation.
Solution Approach 2:
The patent computes geometric features and operators selectively based on the specific needs of each vertex and the complexity of the local geometry. Rather than uniformly applying complex computations to all vertices, the method adapts the level of computational detail to the local requirements, performing partial computations where sufficient and more detailed computations only where geometric complexity demands it, thereby balancing accuracy with computational efficiency.
Data Source
AI summary
Embodiments of the present disclosure provide a method for skinning a character model, a device for skinning a character model, a computer readable medium, and an electronic device, and relate to the field of computer technology. The method comprises: obtaining a first character model and a corresponding skeleton structure; determining a vertex attribute of the first character model and a connecting relationship between the vertices; performing a nonlinear transformation process on the vertex attribute to obtain the first feature, and performing a graph convolution process on the connecting relationship and the first feature to obtain a second feature; determining a global feature and a local feature of the character model according to the second feature; and determining a target feature representing a binding relationship between the first character model and the skeleton hierarchy according to the global feature and the local feature.