A normal distribution transform point cloud registration method based on BFGS correction

doi:10.13474/j.cnki.11-2246.2020.0315

Abstract

Abstract: Point cloud registration is a key problem in point cloud data processing. For the problem of solving the Hessian matrix with high time complexity for the original normal distribution transformation algorithm, a modified normal distribution transform point cloud registration method based on the BFGS algorithm is proposed. The positive definite matrix is updated with the gradient value and incremental parameters of the objective function. The inverse matrix of the Hessian matrix is almost replaced by a positive definite matrix, which reduces the time complexity of the algorithm, ensures that the direction of each iteration of the algorithm is the direction where the function value drops in. The feasibility of this algorithm is verified by simulated data and measured data experiments. This algorithm improves the registration efficiency of the algorithm while maintaining the accuracy of the original normal distribution transformation algorithm.

Key words: point cloud registration, normal distribution transformation algorithm, Hessian matrix, BFGS algorithm, positive definite matrix

CLC Number:

P237

YUAN Zhicong, LU Tieding, LIU Rui. A normal distribution transform point cloud registration method based on BFGS correction[J]. Bulletin of Surveying and Mapping, 2020, 0(10): 38-42.

References

[1] 程效军,贾东峰.海量点云数据处理理论与技术[M].上海:同济大学出版社, 2014.
[2] PAUL J B, NEIL D M. A method for registration of 3D shapes[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1992, 14(2):239-256.
[3] 赵夫群, 周明全. 基于二维图像特征的点云配准方法[J]. 测绘通报, 2017(10):39-42,73.
[4] 李鹏,邢帅,李瑾,等.基于虚拟特征点的三维激光点云粗配准算法[J].地球信息科学学报,2018,20(4):430-439.
[5] 严剑锋,邓喀中.基于特征点提取和匹配的点云配准算法[J].测绘通报,2013(9):62-65.
[6] 武鹏. 平面特征的多站地面激光雷达点云配准[J]. 测绘科学,2016,41(9):108-111.
[7] MAGNUSSON M, LILIENTHAL A, DUCKETT T. Scan registration for autonomous mining vehicles using 3D-NDT[J].Journal of Field Robotics, 2007, 24(10):803-827.
[8] 杨飚, 李三宝, 王力.基于正态分布变换与迭代最近点的快速点云配准算法[J]. 科学技术与程,2017,17(15):91-95.
[9] 陈清艳, 方勇纯,孙宁. 基于曲率特征的改进型3D-NDT配准[C]//第三十四届中国控制会议. 杭州:中国自动化学会控制理论专业委员会,2015.
[10] 胡修祥, 张良. 结合NARF特征的改进型3D-NDT多视点云配准[J]. 信号处理, 2015, 31(12):1674-1679.
[11] 张少坤. NDT算法在古建筑点云数据配准中的应用与研究[J]. 信息化建设,2016(7):291.
[12] 张晓,张爱武,王致华.基于改进正态分布变换算法的点云配准[J].激光与光电子学进展,2014,51(4):100-109.
[13] KAZUNORI H, MARTIAL H, KAZUNORI I. Building 3-D models from unregistered range images[J]. Graphical Models and Image Processing, 1995,57(4):315-333.
[14] VAN T N, TRUNGTHIEN T, et al. 3D point cloud registra-tion based on the vector field representation[C]//20132nd International Association for pattern Recognition Asian Conference on Pattern Recognition. Naha:IEEE, 2013.
[15] 王鹏,李少达,赵雪.基于拟牛顿法改进的3D正态分布变换点云配准算法[J].地理信息世界, 2017, 24(5):42-44.