UAV Building Point Cloud Footprint RegularizationUsing Three-Point Separation Peaks

Abstract

To address the limitations of existing methods in regularizing unmanned aerial vehicle (UAV)-derived building point cloud contours, particularly for complex shapes and noisy data, this study proposes a robust corner points determination algorithm to achieve accurate and adaptable contour regularization. The method in this paper mainly involves two key steps: (1) the initial corner points of boundary points are recognized according to the point-line distance and unbalanced point distribution on both sides of the line; and (2) the cut-off distance of the initial corner points is constructed, and the corner points are further judged based on the decision graph. Simultaneously, two indicators based on cut-off distance and point-line distance deviation metrics are introduced to identify corner points. Comparative analyses showed the proposed method outperformed minimum bounding rectangle, recursive minimum bounding rectangle, multiple-curve growth and improved cubic B-spline fitting methods particularly for irregular contours and noise-affected data. Additionally, the method successfully reconstructed 3D building models by extending roof corner points to the ground.