A New Filter of Morphological Opening by Reconstruction for LiDAR Data

Abstract

In this paper, a new filtering algorithm of LiDAR point clouds is presented, which can work well in complex cityscapes. This method generates marker image from grid DSM by erosion firstly, and then geodesic dilate on marker image repeatedly to implement opening by reconstruction process, finally white top-hat reconstruction are used to achieve the nDSM and classify ground and non-ground object points. When tested against the ISPRS LiDAR reference urban dataset and compared with representative filtering methods, this approach is particularly effective at minimizing Type I error and Total error rates, while maintaining acceptable Type II error rates. So this approach has good reliability and practicability in complex urban area.