Abstract:In the process of point cloud registration,it is easy to produce wrong matching points,long registration time,and low registration accuracy.A registration method is proposed,which bases on 3DSIFT key point detection method combined with BSHOT descriptor to construct a point cloud matching pair.This method firstly uses the differential Gaussian model to detect SIFT key points in the three-dimensional scale space,and secondly constructs a local coordinate system in the neighborhood of the key points to calculate the SHOT descriptor,and converts the SHOT descriptor into a binary descriptor.Then the random sampling consensus algorithm is used to remove the mismatched point cloud,and the transformation matrix of the point cloud is preliminarily estimated.Finally,the ICP algorithm is used to estimate the optimal transformation matrix in the fine registration.Experiments on the data set verify the rapidity of the algorithm in this paper.At the same time,when the overlap rate of the two point clouds is low,the registration accuracy is high.