Slope Adaptive Based Filtering for Airborne LIDAR

In this paper, a three-stage algorithm for raw LIDAR data filtering has been presented. Firstly, the raw data is re-sampled to grid, which size should take into consideration of raw data density and usage intention. Then a slope adaptive region growing is used to segment grid cells to different surface patches. Secondly, segmentation adjacency is established based on topological analysis. Two perceptual cues, that is, connection and elevation are computed for every pair of the adjacent segmentations. By the cues, the upper segmentations are classified as terrain, building, vegetation and others. Finally, the grid cells classified as terrain are used to build approximation of the terrain surface. Raw data are accepted as terrain within a given distance from the surface. To illustrate the effectiveness of the new procedure for DTM production, a series of data sets from ISPRS are tested. The results show that the method can discriminate terrain points from non-terrain points correctly.