A preset fixed value is adopted as the neighborhood radius based on the features of a three-dimensional (3D) point cloud fast point feature histogram (FPFH), resulting in some problems such as arbitrariness, incompleteness, and inefficiency in feature description. Thus, the entire process of point cloud registration becomes less automated and requires more time. To solve these problems, an algorithm is proposed in this study to automatically select the radius of the FPFH neighborhood in 3D point cloud registration. First, the circumferential density of a multipair point cloud was calculated and the maximum circumference radius was retained. Second, the number of iterations was established and the single neighborhood radius was automatically divided according to the number of iterations and the maximum circumference radius of each pair of point cloud. The features of FPFH were extracted based on the divided neighborhood radius and used for registration in the sampling consistency initial registration algorithm. Finally, the circumferential density of the multipair point cloud and the corresponding optimal neighborhood radius were estimated. Further, the mapping function was obtained using the polynomial fitting method; thus, the FPFH feature extraction optimization algorithm was developed. Results show that the proposed algorithm can automatically adapt to the optimal neighborhood radius according to the circumferential density of the point cloud, effectively reduce the incompleteness and redundancy associated with point cloud description, and improve the speed and accuracy of point cloud registration while improving the degree of automation associated with point cloud registration.