Multiple Natural Features Fusion for On-Site Calibration of LiDAR Boresight Angle Misalignment

Boresight angle misalignment is a major error source in a mobile light detection and ranging (LiDAR) system (MLS), which directly affects the overall accuracy and quality of MLS scanned point clouds data. However, the current calibration of the boresight angle misalignment mainly relies on artificial target features or a manual adjustment, and the intensive labors dramatically limit the calibration flexibility. To solve these problems, this article develops a novel on-site calibration method for boresight angle misalignment based on multiple natural features' constraints, which can automatically incorporate multiple natural features extracted from surrounding environments to generate more accurate calibration results for MLS boresight angle without using any artificial targets or specific facilities. First, an improved four-point congruent sets (I-4PCS) algorithm is proposed for registering the MLS point clouds in forward and backward scanned overlapping areas and realizing smooth global registration for point clouds data. Second, a weight principal component analysis (WPCA) approach is presented to automatically extract the appropriate multiple natural features from the well-registered point clouds and establish the appropriate features' representation. Third, according to the extracted multiple features, certain geometric constraints' equations for spherical, linear/cylindrical, and planar features are established based on a model adjustment strategy. Finally, the boresight angle misalignment calibration can be achieved by fitting the corresponding geometric constraints' equations and minimizing the weight through a least-squares adjustment process. The experimental results demonstrate that the proposed method can effectively on-site calibrate the boresight angle misalignment error, and the overall performance of MLS is significantly improved after the calibration based on multiple natural features' constraints.