DETECTION AND ANALYSIS OF GROSS ERROR IN LIDAR POINT
CLOUD BASED ON ROBUST MAHALANOBIS DISTANCE
1) Postgraduate Management Group，Engineering University of CAPF，Xi’an 710086
2)Key Lab of Soil Erosion Process and Control in Loess Plateau，Ministry of Water Resources，Zhengzhou 450003
3) Department of Information Engineering，Engineering University of CAPF，Xi’an 710086
Abstract：Based on the Robust Mahalanobis Distance，an algorithm for detecting and removing gross error in LiDAR point cloud is proposed.First，LiDAR point cloud is divided into several blocks.Then，detMCD (deterministic minimum covariance determinant) algorithm is performed to get robust estimation of the position and scale parameters of each block.With these parameters，Robust Mahalanobis Distance of each point within the block is calculated.As the square of Robust Mahalanobis Distance could be considered to obey Chisquare distribution，the gross error discrimination threshold could be obtained with certain confidence.Finally，gross errors might be detected and removed with the threshold and robust mahalanobis distance in each block.The proposed algorithm was tested in LiDAR data acquired from a typical gully in Qiaozi Valley，Tianshui City of Gansu Province.The algorithm was performed with various combinations of parameters，including the average point number per block in point cloud dividing step and the relative size of subset reserved in detMCD step.The results show that more points were removed as gross errors with increase of the average point number per block，while a larger relative size of subset led to less points detected as gross errors.The comparison of the TIN and profiles generated by LiDAR point cloud before and after the gross error detected and removed shows that the proposed algorithm can remove the gross error in LiDAR point cloud efficiently and achieve higher accuracy and robustness.
Feng Lin,Li Binbing,Huang Lei. DETECTION AND ANALYSIS OF GROSS ERROR IN LIDAR POINT
CLOUD BASED ON ROBUST MAHALANOBIS DISTANCE[J]. jgg, 2014, 34(5): 168-173.
