Effective Selection of Variable Point Neighbourhood for Feature Point Extraction from Aerial Building Point Cloud Data

Existing approaches that extract buildings from point cloud data do not select the appropriate neighbourhood for estimation of normals on individual points. However, success these depends correct normal vector. In most cases, a fixed is selected without considering geometric structure object and distribution input cloud. Thus, heterogeneous cloud, this paper proposes new effective approach selecting minimal neighbourhood, which can vary each point. For point, number neighbouring points are iteratively selected. At iteration, based calculated standard deviation fitted 3D line to points, decision made adaptively about neighbourhood. The make calculation vector accurate. direction then used calculate inside fold feature addition, Euclidean distance mean its boundary context accuracy evaluation, experimental results confirm competitive performance proposed selection over state-of-the-art methods. Based our generated ground truth data, extraction techniques show more than 90% F1-scores.