Abstract: In the paper,a PICP method is proposed to improve an ICP algorithm based on Procrustes analysis,in order to deal with the problems of the traditional ICP algorithm,such as the complication of finding the nearest iteration point,wrong point pairs caused by one way search mechanism,and making the convergence function into local optimum value easily.First,finding the optimal initial transform parameters of point cloud data by comparing the initial transformation parameter and the distance between the two points of iteration point pairs among eight directions of three dimensional space.Second,finding the nearest iteration points with the bidirectional search mechanism and constituting them into a new point cloud data to optimize the ICP algorithm.Finally,solving the least squares function of the cloud data by the Procrustes analysis,then the optimal convergence of ICP registration algorithm can be obtained with higher registration precision.In the experiments,single tooth point cloud data and standard rabbit point cloud data are adopted to do the registrations.It shows that the PICP algorithm can solve the scale-transform and non-uniform point cloud registration problems,with faster convergence and more accurate registration than other methods.Compared with the traditional ICP algorithm,the PICP algorithm shows more benefits in high global convergence,fewer iteration times and strong noise resistance.