Traditional point cloud registration algorithms, such as the Iterative Closest Point (ICP) algorithm, often face challenges like slow convergence, lengthy registration times, and strict initial pose requirements. In light of these issues, a new point cloud registration algorithm is introduced in this study. This innovative approach makes use of the Scale-Invariant Feature Transform (SIFT) for feature point extraction and integrates the Random Sample Consensus (RANSAC) algorithm to eliminate incorrect point pairs. Initially, the SIFT algorithm extracts feature points from both the source and target point clouds, which are then characterized by the Fast Point Feature Histogram (FPFH). Following this, normal vector estimation is applied to identify matching point pairs and filter out erroneous matches using RANSAC. The Sample Consensus Initial Alignment (SAC-IA) algorithm is subsequently utilized to find corresponding point pairs, with the final step involving the ICP algorithm for point cloud registration. The experimental findings reveal that this novel algorithm outperforms the traditional ICP algorithm by reducing registration errors and shortening registration times. Consequently, the proposed approach enhances both the efficiency and accuracy of point cloud registration processes.