The traditional random sampling consistency (RANSAC) algorithm can only perform coarse registration at low efficiency. To address this problem, an improved RANSAC fast point cloud registration algorithm is proposed herein. The proposed algorithm first combines the intrinsic shape signatures and fast point feature histogram algorithms to obtain feature descriptors and then employs pre-estimation and three-dimensional (3D) grid segmentation to improve the RANSAC algorithm. Finally, it is compared with the traditional sample consensus initial alignment algorithm. Our experimental results demonstrate that the proposed algorithm can quickly and accurately eliminate false matching points and solve the affine transformation matrix without secondary registration. In comparison with the traditional registration algorithm, the proposed algorithm demonstrates good robustness in large-scale 3D point cloud registration and significantly improves the registration efficiency while ensuring accuracy.