Revised progressive morphological method for ground point classification of airborne LiDAR data

Abstract

Airborne Light Detection and Ranging (LiDAR) has been very effectively used in collecting terrain information over different scales of area. Inevitably, filtering the non-ground returns is the major step of digital terrain model (DTM) generation and this step poses the greatest challenge especially for tropical forest environment which consists of steep undulating terrain and mostly covered by a relatively thick canopy density. The aim of this research is to assess the performance of the Progressive Morphological (PM) algorithm after the implementation of local slope value in the ground filtering process. The improvement on the PM filtering method was done by employing local slope values obtained either using initial filtering of airborne LiDAR data or ground survey data. The filtering process has been performed with recursive mode and it stops after the results of the filtering does not show any improvement and the DTM error larger than the previous iteration. The revised PM filtering method has decreasing pattern of DTM error with increasing filtering iterations with minimum ±0.520 m of RMSE value. The results also suggest that spatially distributed slope value applied in PM filtering algorithm either from LiDAR ground points or ground survey data is capable in preserving discontinuities of terrain and correctly remove non-terrain points especially in steep area.