摘要
针对动画角色模型由于姿态变化所导致的分割不一致问题,提出语义本体驱动的层次一致性分割算法.针对三维模型提取出语义标签和局部几何特征之间的映射关系,形成分割本体.在分割过程中,采用支持向量机(SVM)和局部几何特征识别出语义标签,并根据语义标签驱动层次进行分割,保证动画角色模型分割层次的一致性.针对姿态变化下分割边界所具有的等周长性,采用泊松方程定义等值线.此优化方法使分割边界光滑,还使其在姿态变化下具有一致性.在实验部分,对不同姿态下的各类动画角色模型进行验证分析,得到一致的层次分割效果.与已有方法比较,所提出的分割本体能够解决不同类模型优化参数的自适应选择问题,提高了分割质量.
A segmentation algorithm was proposed based on semantic ontology to solve the problem that animation character model is inconsistent because of the changing of poses.The segmentation ontology was constructed according to the map between semantic label and geometrical characteristic.In the process of segmentation,the semantic label of segmented parts was recognized using support vector machine（SVM）and geometrical characteristic.In this way,the hierarchical segmentation of dynamic models could be consistent with the help of semantic label.In addition,the length of boundary was almost equal under different poses,therefore,segmenting boundary was refined by Poisson equation.The refined method made the animation character models smooth and consistent.In experiment,the segmenting result of the tested dynamic models show that the proposed algorithm is very stable.Compared with the existing methods,segmentation ontology can select optimizing parameters for different 3Dshapes.As a result,the proposed method can achieve boundary of segmenting quality over the exiting methods.
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2015年第9期1634-1641,共8页
Journal of Zhejiang University：Engineering Science
基金
国家自然科学基金资助项目(61272304)
浙江省自然科学基金资助项目(LY15F020024)
浙江省文物局资助项目(2014014)
关键词
动画角色模型
分割本体
层次分割
泊松方程
支持向量机(SVM)
animation character model
segmenting ontology
hierarchical segmentation
Poisson equation
support vector machine（SVM）
