科研之友
微信
新浪微博
Facebook
分享链接
摘要
In the processing of airborne light detection and ranging (lidar) point clouds, filtering is one of the core steps in its applications, such as digital elevation model (DEM) generation. The classic progressive triangulated irregular network densification (PTD) has been proved to be effective in filtering, but this method is sensitive to maximum angle and maximum distance, which leads to misclassification in filtering. In this article, we analyse the connection between the slope and those two key parameters and propose a novel parameter-free PTD (PFPTD) algorithm. In the PFPTD algorithm, slope is predicted through Kriging, and the predicted slopes are embedded into iterative densification of unlabelled points. To test the performance of the proposed algorithm, seven benchmark data sets provided by the International Society for Photogrammetry and Remote Sensing Working Group III/3 are employed. Among the 7 data sets, 15 reference sub-area samples with manual filtering results are utilized for quantitative analysis. Experiment results suggest that the proposed algorithm is capable of improving the performance of filtering while demanding less involvement in parameter selection, which is significantly important in automatic and high-accuracy DEM generation.
