基于隧道变形测量中的点云粗配准实验研究

Experimental Study on Coarse Registration of Point Cloud in Tunnel Deformation Measurement

三维激光隧道点云配准精度直接影响到隧道变形分析的可靠性。点云配准精度很大程度上依赖于点云初始姿态,为获取更好的初始点云姿态,分析了不同全局配准算法的优劣,并通过实验得出不同全局配准算法应用于隧道点云模型的粗配准结果。根据实验表明,全局配准算法(FGR)算法在配准精度上与传统全局配准算法(RANSAC)近接,而在算法耗时、稳健性等方面明显优于经典的RANSAC算法,说明FRG算法更适合于隧道点云数据的初始姿态配准。

The accuracy of 3d laser tunnel point cloud registration directly affects the reliability of tunnel deformation analysis. The point cloud registration accuracy largely depends on the initial point cloud attitude. To obtain a better initial point cloud attitude, the advantages and disadvantages of different global registration algorithms are analyzed, and the rough registration results of dif-ferent global registration algorithms applied to the tunnel point cloud model are obtained through experiments. According to the experimental results, the global registration algorithm (FGR) is close to the traditional global registration algorithm (RANSAC) in terms of registration accuracy, and is better than the classical RANSAC algorithm in terms of algorithm time robustness, indicating that the FRG algorithm is more suitable for the initial attitude registration of tunnel point cloud data.