To address issues such as mismatches, low registration accuracy, and low efficiency during the point cloud registration process, this paper presents a point cloud registration algorithm based on local curvature and distance features. The algorithm extracts key points from point cloud data using local curvature information and the weighted distance from center points to their neighbors. Firstly, the K-4PCS algorithm was utilized for point cloud coarse registration, then those key points extracted from a local point cloud using linear least squares algorithm were used to complete precise registration using surface ICP algorithm. The method is tested on multiple sets of point cloud data and the results show that the key point extraction method improves the feature of key points compared to the ISS algorithm and the SIFT algorithm, providing a better basis for subsequent registration operations. In subsequent registration, under the same conditions, the algorithms proposed in this paper has improved efficiency, accuracy, and accuracy by approximately 56% and 57% compared to other algorithms in coarse registration. In precision registration, the efficiency has been improved by about 60%. This method can serve as a reference for point cloud registration scenarios where both efficiency and accuracy are critical.