Production of Digital Terrain Model (DTM) in Dense Forest Areas with Combined Airborne LiDAR Data Filtering Algorithms

Producing digital elevation models (DTMs) for forested areas is a challenging task. Nowadays, because of the capability of LiDAR pulses to penetrate through the vegetation canopy, airborne laser scanner technology has attracted enormous interest for DEM extraction especially in forested terrain. Producing Digital Terrain Models from airborne LiDAR data consist of two main steps of filtering and interpolation. Several automatic filtering algorithms for laser scanning data have been proposed. The capability of these algorithms to separate ground and non-ground points is not the same. The aim of this study is to combine two algorithms to achieve higher accuracy. In order to separate ground from non-ground points at first, noise was removed from LiDAR data, then the third moment (skewness) was calculated and finally slope-based algorithm was applied. The skew amount of LiDAR data was equal to 0.48; so the points with greater height were recognized as object points and removed from point clouds. Some of the non-ground points were recognized by mistake as the ground points which were removed by applying slope-based algorithm. Slope-based algorithm with five threshold slopes 18, 20, 22, 24 and 26 degrees was applied on point clouds and after calculating type I, type II and total errors the best threshold slope of 22 degrees was found and implemented. Finally, a digital terrain model was produced by B-spline interpolation method. The results showed 21 cm vertical accuracy for produced digital elevation model which comparing to previous works was acceptable. User's and producer's accuracy for ground points was 93.38 and 94.5 and for non-ground points was 94.5 and 95.2, respectively. This accuracy indicates that the ground and non-ground points were separated very carefully.