A wavelet based algorithm for DTM extraction from airborne laser scanning data

Abstract

The automatic extraction of Digital Terrain Model (DTM) from point clouds acquired by airborne laser scanning (ALS) equipment remains a problem in ALS data filtering nowadays. Many filter algorithms have been developed to remove object points and outliers, and to extract DTM automatically. However, it is difficult to filter in areas where few points have identical morphological or geological features that can present the bare earth. Especially in sloped terrain covered by dense vegetation, points representing bare earth are often identified as noisy data below ground. To extract terrain surface in these areas, a new algorithm is proposed. First, the point clouds are cut into profiles based on a scan line segmentation algorithm. In each profile, a 1D filtering procedure is determined from the wavelet theory, which is superior in detecting high frequency discontinuities. After combining profiles from different directions, an interpolated grid data representing DTM is generated. In order to evaluate the performance of this new approach, we applied it to the data set used in the ISPRS filter test in 2003. 2 samples containing mostly vegetation on slopes have been processed by the proposed algorithm. It can be seen that it filtered most of the objects like vegetation and buildings in sloped area, and smoothed the hilly mountain to be more close to its real terrain surface.