Aiming at the problems of long reconstruction time，slow convergence and error matching corresponding points for the point cloud registration, a new algorithm based on the intrinsic shape signature (ISS) feature points combined with the improved iterative closest point (ICP) algorithm is proposed. Firstly, the feature points of point cloud are extracted by the ISS algorithm and described by the fast point feature histograms algorithm. Then, the initial registration of point cloud is completed by using the sample consensus initial alignment algorithm to make the two different angle point clouds obtain a relatively good initial position. Finally, the ICP registration efficiency is promoted by the k-dimension tree nearest neighbor search algorithm. The experimental results show that the proposed algorithm has higher registration accuracy and faster execution speed than the traditional registration algorithms.