Intelligent optimization of seam-line finding for orthophoto mosaicking with LiDAR point clouds

A detailed study was carried out to find optimal seam-lines for mosaicking of images acquired by an airborne light detection and ranging (LiDAR) system. High ground objects labeled as obstacles can be identified by delineating black holes from filtered point clouds obtained by filtering the raw laser scanning dataset. An innovative A* algorithm is proposed that can automatically make the seam-lines keep away from these obstacles in the registered images. This method can intelligently optimize the selection of seam-lines and improve the quality of orthophotos. A simulated grid image was first used to analyze the effect of different heuristic functions on path planning. Three subsets of LiDAR data from Xiuan, Dunhuang, and Changyang in Northwest China were obtained. A quantitative method including pixel intensity, hue, and texture was used. With our proposed method, 9.4%, 8.7%, and 9.8% improvements were achieved in Dunhuang, Xiuan, and Changyang, respectively.