A large-scale multiview point cloud registration method based on distance statistical distribution of weak features.

There are issues such as poor layered smoothness, model distortion, and error accumulation in the process of large-scale weak feature point cloud stitching and registration. This paper proposes a precise point cloud registration method based on distance statistical distribution. By summarizing and statistically analyzing the distance thresholds during the iterative process, it accurately determines the closest points for point cloud registration, thus avoiding the problem of low registration accuracy caused by manually setting distance thresholds. By statistically analyzing the distance intervals of corresponding points in the point cloud to eliminate erroneous correspondences and utilizing pose graph optimization for global pose, this method ensures the smoothness and accuracy of point cloud registration. Experiments have validated the effectiveness of this method. Comparative experiments have demonstrated that this method surpasses traditional point cloud registration methods in terms of accuracy, convergence, and robustness. [ABSTRACT FROM AUTHOR]