关键词: 机载光探测与测量（LiDAR）, 滤波, 偏度, 阈值, 非监督分类

Abstract: Generally, filtering for point clouds is considered as a primary step for airborne LiDAR data post-processing. However, it is still under development. By reviewing and summarizing the existing filtering algorithms, the concepts of skewness and kurtosis in statistics are introduced to the framework, and a novel unsupervised classification algorithm to differentiate ground and non-ground points based on skewness balancing is presented. Digital Terrain Model（DTM） is effectively extracted from Digital Surface Model（DSM） generated from LiDAR data by exploiting the statistical moments principle. As an alternative to traditional approaches, the ultimate advantages of this algorithm are parameter-and threshold-freedom and independence from LiDAR data format and resolution. Experiment results show that the algorithm is highly practicable and adaptive, and can meet the required precision properly.

Key words: airborne Light Detection And Ranging（LiDAR）, filtering, skewness, threshold, unsupervised classification