基于Laplace谱嵌入和Mean Shift的 三角网格一致性分割

针对现有网格分割算法对模型姿态及噪声敏感的不足,提出一种基于Laplace谱嵌入和Mean Shift聚类的网格一致性分割算法。采用Laplace-Beltrami算子,将3维空域中的网格模型转化成高维Laplace谱域中的标准型,降低了姿态变化和噪声对分割算法的影响,并增强了网格的结构可分性;在高维谱域中,采用非参数核聚类MeanShift算法,获取模型有视觉意义的语义区域。实验结果表明:该算法可以快速有效地实现具有分支结构三角网格模型的有意义分割且对模型姿态和噪声具有较好的鲁棒性。

Consistence segmentation of triangle mesh using Laplace spectral embedding and Mean Shift

In order to overcome the disadvantage of being sensitive to model gesture and noise in the present mesh segmentation algorithms,we present a consistent mesh segmentation algorithm based on Laplace spectral embedding and Mean Shift. We convert mesh into a normal form from the space domain to the spectral domain by using the Laplace-Beltrami operator. The noise is suppressed and spectral embedding enhances the structural segmentability. We adopt Mean Shift,a nonparametric kernel clustering technique,to gain the visual meaningful semantic patch or sub-mesh in the spectral domain. The experiment results show that the proposed algorithm can yield meaningful result rapidly and effectively for meshes which has an evident branch structure.Meanwhile,this approach is invariant to pose of model and robust to noise.