Application of Intelligent Interpolation Methods for DTM Generation of Forest Areas Based on LiDAR Data

The increased ability of laser pulses to penetrate the vegetation cover has made LiDAR technology an attractive option for the generation of Digital Terrain Models (DTM) of forest areas. A LiDAR-based DTM generation consists of two phases, filtering and interpolation. This study suggests a combination of a slope-based and a hybrid filter for filtering of raw LiDAR data with innovative interpolation methods including artificial neural network (ANN) and artificial emotional neural network (ENN). Also, the genetic algorithm (GA) is used to improve both polynomial and inverse distance weighting (IDW) interpolation methods. The performance of those intelligent methods was also compared with that of conventional methods such as kriging and radial basis functions (RBF). For practical applicability of the presented method, two LiDAR datasets of forest regions in the Golestan province of Iran were selected, one with dense vegetation cover (Tavar-kuh) and one with reduced cover density (Shastkola-River basin). The results of these studies indicate that the hybrid filter outperforms the slope-based filter. Also the conventional interpolation methods were unable to achieve the accuracies offered by the intelligent methods. The elevation of the result shows that the best DTM reaches a root-mean-square error (RMSE) of 0.09 m for Tavar-kuh and 0.12 m for Shastkola-River basin by hybrid filter with an ENN interpolation.