Automatic Building Boundary Extraction from Airborne LiDAR Data Robust to Density Variation

Abstract

The alpha-shape ( $\alpha $ -shape) concept, which has its origin in computational geometry, is usually applied in building boundary extraction from airborne LiDAR data. However, the results depend on the appropriate choice of the parameter $\alpha $. Despite several studies in the literature, the adaptive choice of the parameter $\alpha $ persists a challenge in boundary extraction, especially when abrupt density variations occur. To overcome this limitation, this letter proposes a new approach combining five estimation strategies. In the proposed method, these strategies are tested sequentially, prioritizing the one that provides greater level of details. The experiments were conducted considering buildings with different characteristics, which were selected from two LiDAR data sets with the average point densities of 12 points/m2 and 4 points/m2. The obtained results, presenting $\boldsymbol {F} _{{\text {score}}}$ and PoLiS around 98% and 0.32 m, respectively, indicate the robustness of the proposed method even when abrupt density variation occurs.