处理多流形聚类问题，从相似性矩阵的角度出发，充分利用流形采样点所内含的自然的局部几何结构信息来辅助构造更合适的相似性矩阵并进而发现正确的流形聚类。-Handle multiple manifolds clustering problem,　the perspective of the similarity matrix starting, make full use of local natural manifold geometry information contained sampling points to assist construct a more suitable similarity matrix and thus find the right manifold Clustering .