基于凹凸特征的复杂岩石体点云配准算法研究与应用OA

This paper introduces an advanced point cloud registration algorithm for rock masses,addressing challenges such as low overlap rates and complex feature distributions in complex geological environments.The method employs the Robust Neighbor Reweighted Local Centroid(RNRLC)technique,combined with a classification mechanism,to extract concave-convex feature points and construct descriptors based on spatial distance and concave-convex information.An optimized clustering mechanism enhances descriptor reliability while suppressing noise and local anomalies.Correspondences are identified via edit distance,and congruent triangles are constructed to compute the transformation matrix.A two-stage dynamic parameter adjustment ensures robust alignment,applying the transformation to the entire point cloud.Experiments show the proposed method outperforms traditional algorithms,such as ICP and CCEE,in scenarios with low overlap and large angular deviations,offering a reliable solution for point cloud registration in complex geological settings.