Accuracy assessment of LIDAR-derived elevation value over vegetated terrain in Tropical region

Abstract

Airborne LiDAR has been widely used to generate good quality of Digital Terrain Model (DTM). Normally, good quality of DTM would require high density and quality of airborne LiDAR data acquisition which increase the cost and processing time. This study focuses on investigating the capability of low density airborne LiDAR data captured by the Riegl system mounted on an aircraft. The LiDAR data sampling densities is about 2.2 points per m2. The study area is covered by rubber trees with moderately dense understorey vegetation and mixed forest. The ground filtering procedure employs the adaptive triangulation irregular network (ATIN) technique. A reference DTM is generated using 76 ground reference points collected using total station. Based on this DTM the study area is divided into different classes of terrain slopes. The point clouds belong to non-terrain features are then used to calculate the relative percentage of crown cover. The overall root mean square error (RMSE) of elevation values obtained from airborne LiDAR data is 0.611 m. The slope of the study area is divided into class-1 (0-5 degrees), class-2 (5-10 degrees), class-3 (10-15 degrees) and class-4 (15-20 degrees). The results show that the slope class has high correlation (0.916) with the RMSE of the LiDAR ground points. The percentage of crown cover is divided into class-1 (60-70%), class-2 (70-80%), class-3 (80-90%) and class-4 (90-100%). The correlation between percentage of crown cover and RMSE of the LiDAR ground points is slightly lower than the slope class with the correlation coefficient of 0.663.