The main solution for large-scale point clouds registration is to first obtaina set of matched3D keypoint pairs and then accomplish the point cloud registration task based on these matched keypoint pa

In summary, this study proposes a deep learning-based approach for 3D keypoints detection in large-scale point clouds registration. The proposed method achieves better inlier ratios and state-of-the-art registration performance compared to existing methods. This approach has the potential to improve the accuracy and efficiency of large-scale point clouds registration tasks in various applications, such as robotics, autonomous driving, and 3D reconstruction.