Plan3D: Viewpoint and Trajectory Optimization for Aerial Multi-View Stereo Reconstruction
share
We introduce a new method that efficiently computes a set of rich viewpoints and trajectories for high-quality 3D reconstructions in outdoor environments. The input images of the reconstruction are taken with a commodity RGB camera which is mounted on an autonomously navigated quadcopter, and the obtained recordings are fed into a multi-view stereo reconstruction pipeline that produces high-quality results but is computationally expensive. Our goal is to automatically explore an unknown area, and obtain a complete 3D scan of a region of interest (e.g., a large building). In this process, the scan is constraint by the restricted flight time of quadcopters and the heavy compute costs of the subsequent 3D reconstruction -- i.e., only a small number of images can be recorded and processed. To this end, we introduce a novel optimization strategy that respects these constraints by maximizing the information gain from sparsely-sampled view points while limiting the total number of captured images. The core of this strategy is based on the concept of tri-state space classification, which is common in volumetric fusion approaches, and includes labels for unknown, free, and occupied space. Our optimization leverages a hierarchical and sparse volumetric data structure that takes advantage of the implicit representation, where its main objective is to convert unknown space into known regions. In addition to the surface geometry, we utilize the free-space information to avoid obstacles and determine feasible flight paths. A simple tool can be used to specify the region of interest and to plan trajectories. We demonstrate our method by obtaining a number of compelling 3D reconstructions, and provide a thorough quantitative evaluation for our optimization strategy.
READ FULL TEXT