Building detection from lidar data using entropy and the k-means concept

Abstract

Information obtained from LiDAR data processing is considered in a variety of applications, among them urban planning. In this context, buildings play a substantial role, since a high percentage of the urban landscape is occupied by them. In the literature, many methodologies have been developed aiming at the detection of building using remote sensing data. The approaches can be developed by applying different ideas: regularity of cluster boundary, plane fitting, radiometric data and also in geometric attribute derived from LiDAR. This paper proposes a method of building detection based on the use of the entropy concept and the K-means algorithm in which the training step is dispensed with. The experiments were performed considering two LiDAR datasets with different densities (12.5 pts/m2 and 4 pts/m2). Visual and qualitative analysis enabled verification of the potential of the proposed method, which presented satisfactory results for both datasets.