INTEGRATION OF DIFFERENT FILTER ALGORITHMS FOR IMPROVING THE GROUND SURFACE EXTRACTION FROM AIRBORNE LIDAR DATA
An important step for processing airborne Light Detection And Ranging (LiDAR) data is point cloud filtering. Points striking on vegetation and man-made objects and low points (points significantly lower than neighboring points) are filtered out, leaving ground points for generation of digital terrain models (DTM). A variety of filter algorithms have been developed, which have disparate performance in different landscape and environment. This study investigates the potential of integrating the results of different filter algorithms for improving the ground surface extraction from the LiDAR point cloud. A simple procedure was proposed based on a statistical approach to identify and remove filtering errors and combine ground points from each filtering result. The procedure was tested in an area with rugged terrain covered by dense vegetation of variable heights. The filtering results of two popular filter algorithms, progressive TIN (Triangulated Irregular Network) densification and hierarchical robust interpolation, were integrated. The filtering results of two algorithms and the integration result were qualitatively evaluated. The evaluation results indicated that the proposed integration procedure can remove most vegetation points that were not filtered out by filter algorithms, and combine ground points from each filtering result.