Filtering airborne Lidar data with morphological analysis methods

A multi-scale solution was proposed for synchronous implementation of ground point identification (point clouds filtering) and morphological feature analysis, to adapt to various landscapes. First, a pseudo-grid and hybrid-interpolation methods were employed for effective operation of the point clouds. Then, the morphological analysis, including the background estimation, object scale analysis and object direction analysis, was implemented together with an edge-based morphological filter in a same multi-scale framework. The morphological analysis also helps estimate parameters for the filter. Lastly, a second filtering process was followed to identify ground points from residual points, based on the local landscape features. The tests with the benchmark dataset from international society for photogrammetry and remote sensing (ISPRS) show that the morphological analysis is rational and the average filtering accuracy is better than 90% for different landscapes, especially in the complex landscape with a mixture of buildings and vegetation on the slope terrain.