DTM Generation in Forested Areas from Full-Waveform Airborne LiDAR Data

This study was aimed at improving overall quality of Digital Terrain Model (DTM) extraction from full-waveform LiDAR data. Specifically, the primary goal was to develop a novel method to improve DTM extraction by utilizing low amplitude pulses that are generated by terrain under vegetation, but undetectable using traditional Gaussian decomposition techniques. The secondary objective was to validate the developed methodology using ground reference data.

An integrated approach was developed to detect weak returns backscattered by the bare terrain using full-waveform data and implemented using Microsoft Visual Studio. In this approach, echo detection, identification of terrain points, and generation of the triangulated irregular network (TIN) were iteratively carried out. To validate the proposed method, airborne LiDAR datasets obtained from a Riegl’s LMS-Q560 over five study sites in the Great Lakes-St. Lawrence forest region near Sault Ste. Marie, Ontario, Canada were used. The generated DTMs were compared with those obtained from the commercial software, TerraSolid’s TerraScan, based on ground measurements. The validation results show that using the developed method, the improvement in DTM was up to 21% for the five study areas, but up to 29% only considering heavily wooded areas with variable terrain. In addition, the developed methodology demonstrated an increase in LiDAR density and coverage of terrain points detected (up to 10-15%), when compared to TerraScan’s ground extraction routine.