Segmenting and Estimating Multiple 3-D Rigid Motions From 2-D Point Correspondences
用二维点对应数据分割和估计多刚体运动

Given the 2-D point correspondence set, which may be corrupted by both Gaussian noise and outliers, of a scene containing multiple rigid objects, and based on the rigidity constraint, the proposed algorithm, consisting of the initial partial match generation and growth, can segment the above set into several subsets corresponding to different 3-D rigid motions and outliers. The motion parameters of each rigid motion can then be obtained by the motion estimation algorithm for single 3-D rigid object. Computer simulation results demonstrate the effectiveness and efficiency of the algorithm.