Point cloud registration algorithms baseFd on feature descriptions rely heavily on the performance of their feature descriptors. However, classical local feature descriptors often suffer from insufficient robustness and speed. To address these issues, this paper introduces a novel point cloud registration method utilizing the Space Angle and Curvature (SAAC) local feature descriptor. The SAAC descriptor vector captures local features of target points, and then a feature screening process determines accurate matching pairs. Further, the final registration is achieved through Singular Value Decomposition (SVD). Finally, experimental evaluations on public datasets show that the SAAC descriptor outperforms traditional descriptors in terms of robustness and speed, while the resulting point cloud registration method exhibits significant advantages.