针对当前三维骨架提取方法复杂度较高、提取结果不够准确，以及专门针对人体模型的方法较少等问题，提出一种基于模型分割的三维人体骨架提取方法。首先，根据模型顶点与末端特征点的最小测地距离将模型分割；然后由归一化的测地距离函数确定模型各顶点所属拓扑层次；接着在模型分割的基础上依据拓扑层次提取出原始骨架点；最后经过微调，将各骨架点按照拓扑关系连接得到较为精确的人体骨架。实验结果表明，该方法有效降低了骨架提取算法的复杂度，且对不同姿势的人体模型均可获得较为准确的提取结果。%Aiming at the problems that current 3D skeleton extraction methods are highly complicated,the extraction results are not accurate enough,and few of them are specifically for human body models,we proposed a model segmentation-based 3D human body skeletons extraction method.First,we divided the model into parts according to the minimum geodesic distances between feature points of its vertices and ends.Secondly,we determined the topological hierarchies of each vertex of the model by the normalised geodesic distance function. Thirdly,based on model segmentation we extracted the original skeleton points according to their topological hierarchies.Finally,after some fine tuning,we connected all the skeleton points according to their topological relations and obtained a quite accurate human skeleton. Experimental results showed that the method reduces the complexity of the skeleton extraction algorithms effectively and gets more accurate extraction result for human body models with different postures.
展开▼
