A Multi-scale Progressive Framework for Ground Segmentation of Airborne LiDAR Data

Abstract

LiDAR technology can be used to carry out ground extraction with higher resolutions. The technique can penetrate the topping of vegetation and collect ground points for bare land estimation. But for thick forest, the simple selection of return echoes to generate ground is not accurate. To solve the problem, this paper presents a progressive framework to segment ground out of densely-vegetated area from LiDAR data. A morphological filter is applied to the process by a multi-scale representation which can adaptively respond to different environment with the parameter of the structure element is progressively increased by one pixel from the starting value at each step. The proposed method also incorporates image inpainting in the implementation of a smooth initial data grid and missing point data restoration. After all, the performances of the framework are tested against different data of densely-vegetated and sloped terrain, and yield a better result.KeywordsAirborne LiDARground segmentationmulti-scale representation