Topology-based UAV path planning for multi-view stereo 3D reconstruction of complex structures

Abstract This paper introduces a new UAV path planning method for creating high-quality 3D reconstruction models of large and complex structures. The core the is incorporating topology information surveyed structure to decompose multi-view stereo into collection overlapped view optimization problems that can be processed in parallel. Different from existing state-of-the-arts recursively select vantage camera views, iteratively resamples all nearby cameras (i.e., positions/orientations) together achieves substantial reduction computation cost while improving quality. approach also provides higher-level automation function facilitates field implementations by eliminating need redundant initialization as studies. Validations are provided measuring variance between reconstructions ground truth models. Results three synthetic case studies one real-world application presented demonstrate improved performance. expected instrumental expanding adoption UAV-based