Comprehensive Analysis of Dense Point Cloud Filtering Algorithm for Eliminating Non-Ground Features

Point cloud and LiDAR Filtering is removing non-ground features from digital surface model (DSM) and reaching the bare earth and DTM extraction. Various methods have been proposed by different researchers to distinguish between ground and non- ground in points cloud and LiDAR data. Most fully automated methods have a common disadvantage, and they are only effective for a particular type of surface. Also, most of these algorithms have good outcomes in simple landscapes and not suitable in complex scene. In this article, the filtering methods are divided into three groups: First: traditional methods including slope-based methods, surface-based methods, morphology methods, TIN-based method, segmentation methods and other rule based filtering methods, second: methods that have specific algorithms or improved efficiency of existing algorithms and finally third filtering techniques: based on new machine learning and deep learning techniques. Then investigate and analysis comprehensively the operational problems, their challenges and efficiency of this methods for different areas mountain, forest, urban. Identify and advantages and disadvantages of each method and suggestions for using different methods in different areas is presented. The results of this analysis indicate that the combination of improved and new methods of machine learning and deep learning are suggested in order to improve the performance of filtering techniques.