Validation of Island 3D-mapping Based on UAV Spatial Point Cloud Optimization: a Case Study in Dongluo Island of China

Abstract

The unmanned aerial vehicle (UAV) remote sensing is of small volume, low cost, fine timeliness, and high spatial resolution, and has the special advantage on island surveying. Focus on the inaccurate elevation of non-ground point cloud without lidar device, this study explored a methodology for island three-dimensional (3D) mapping and modelling based on spatial point clouds optimization with a K-Nearest Neighbors Adaptive Inverse Distance Weighted (K-AIDW) interpolation algorithm. By classifying the UAV point clouds into ground, vegatetation, and structure, the K-AIDW algorithm was applied to optimize the elevations of non-ground point clouds (vegetation and structure) to recalculate Z values. The aerophotogrammetry result was generated based on the optimized spatial point clouds. Finally, the 3D model of Dongluo Island was reconstructed and rendered in Metashape. The accuracy evaluation result shows that the max-errors of ground control points (–0.0154 in X, 0.0305 in Y, and 0.0133 in Z) and the checkpoints (–0.091 in X, –0.176 in Y, and 0.338 in Z) can meet the error-tolerance requirements of the corresponding terrain on the 1:500 scale set by the national standard of GB/T 23236-2009 in China. It is found that the K-AIDW algorithm displayed the best Z accuracy (root-mean-square error of 0.2538) compared with IDW (0.3668) and no-optimized (1.6012), proving it is an effective methodology for improving 3D-modelling accuracy of island.