基于特征增强与极大簇优化的点云高精度配准

打开文本图片集

中图分类号：TD-0 文献标志码：A 文章编号：1672-1098（2025）06-0069-07

Abstract： Objective To addressthe challenges oflowaccuracy，poor efficiency，and limited adaptability to lowoverlap scenarios in point cloud registration under complex geological structures.Methods this paper proposes an optimized registration method that combines the Neighborhood-weighted Regional Local Centroid algorithm （NRLC） with the Maximal Clique （MAC） method，referred to as NRLC-MAC. The method first employs the NRLC algorithm to extract multi-scale geometric features from the point cloud， including concave-convex and boundary features.By introducing distance weights and shape variation weights， the feature vectors are optimized， significantly improving the distinctivenessof feature points.Additionally， the data volume is reduced by over 90% ， leading to a substantial improvement in computational efficiency. Subsequently， the MAC method is applied to construct a compatibility graph in graph space. By employing maximal clique search and hypothesis evaluation strategies，the optimal correspondences among feature points are identified，enabling high-precision point cloud registration. Experiments were conducted using real tunnel engineering point clouds and the public RockBench rock dataset. Results The results show that the proposed method maintains high registration accuracy even under low-overlap conditions，with the root mean square error （RMSE）of tunnel point cloud registration reaching 1.652×10-1m and the RMSE for rock point clouds reaching 4.652×10-2m .Meanwhile， the data volumes werereduced by over 99% and 90% ，respectively， significantly lowering computational resource consumption. Conclusion The NRLC-MAC method demonstrates strong robustnessand applicability in the registration of point clouds from complex geological structures，ofering reliable technical support for 3D geological reconstruction and stability analysis.

Key words： point cloud registration; complex geological structures; feature extraction; NRLC-MAC; low overlap ratio

点云配准是地质工程、数字孪生和地下空间建模中的关键技术，在隧道监测、岩体稳定性分析及三维重建等领域具有重要应用价值[。（剩余9539字）