This study introduces an enhanced point cloud registration method to rectify the low registration precision and need for a superior initial input transformation matrix associated with traditional Iterative Nearest Point cloud registration methods. The improved approach integrates block concepts and point cloud plane fitting for feature point extraction. Firstly, the point cloud is partitioned, and the plane fitting and feature point extraction of the block point cloud are carried out by combining the sampling consistency theorem. Secondly, the fast point pair histogram is used to describe the characteristics of point cloud, and the sampling consistency registration algorithm is used for initial registration, which lays a good foundation for fine registration. Precise registration accelerates the iterative closest point algorithm through K-D tree improvement to realize the overall registration of point clouds. The experimental results show that the proposed point cloud registration method improves the registration accuracy by 89.23% and 31.45% compared with the traditional nearest iteration point method on the two test sets. Compared with other point cloud registration methods, the proposed method also has certain advantages.