Experimental evaluation of ALS point cloud ground extraction tools over different terrain slope and land-cover types

The article presents an evaluation of different terrain point extraction algorithms for airborne laser scanning (ALS) point clouds. The research area covers eight test sites with varying point densities in the range 3–15 points m⁻² and different surface topography as well as land-cover characteristics. In this article, existing implementations of algorithms were considered. Approaches that are based on mathematical morphology, progressive densification, robust surface interpolation, and segmentation are compared. The results are described based on qualitative and quantitative analyses. A quantification of the qualitative analyses is presented and applied to the data sets in this example. The achieved results show that the analysed algorithms give classification accuracy depending on the landscape and land cover. Although the results for flat and mountainous areas as well as for sparse and dense vegetation are in line with previous tests, this analysis provides an overview of situations in which the quantitative evaluation is not enough to correctly assess the classification results.