A new method for estimating building dominant directions in LiDAR data based on straight skeleton

Building boundary extraction is a necessary progress in many applications such as urban mapping, flood management, 3D city model reconstruction etc. Plenty number of researches have shown that dominant directions estimation is an essential step in building boundary extraction and regularization. This study presents an approach to giving approximate dominant directions of buildings from raw LiDAR data which based on straight skeleton analysis of simple polygon. The process consists of a sequence of three steps: trace and coarsely simplify building boundary polygon; generate straight skeleton of the boundary polygon; estimate dominant directions based angle and length threshold and a least square solution. The derived dominant direction can be utilized for a progress of building boundary regularization. Efficiency of the algorithm is proved by experiments on two datasets and the results are evaluated.