UAV-Based SLAM and 3D Reconstruction System

3D reconstructing a landscape is a prevalent problem that attracts a lot of interest in recent years. This project intended to verify whether the hypothesis of a UAV-based SLAM and 3D reconstruction system is practical. A GPS-Fused SLAM system is built based on ORB-SLAM. Inverse depth is also implemented to make the system suitable for a UAV-based platform. Meanwhile, REMODE is a depth filter and is tested as not being well enough as a dense mapping module. In the end, PMVS is implemented to build a dense map of the environment which produces a reasonable result. The small-scale-scene experiments produce the total error ratio of 5.60% in the x-y plane and 6.59% in the z axis.