Multi-Dimensional Mapping, Tracking and Classification using Inertial/Vision/GPS on a UAV
Authors: Salah Sukkarieh, Mitch Bryson, Ali Goktogan.
In this talk we will present algorithms for the fusion of monocular vision-based SLAM with GPS to remove ambiguities in translation, rotation and scale of a constructed 3D map and large mapping uncertainty during high dynamic motion of the camera. The approach provides the advantages of a low-cost, low-weight system, which is ideal for airborne mapping applications (the subject of this talk).
When accurate 3D reconstruction is achieved, this information can furthermore be used to classify objects within the terrain for missions such as invasive plant and weed detection, biomass mapping and animal counting.
We will also discuss processes in information visualisation for displaying the underlying complex and dynamic data structures in the mapping/classification/tracking process, for the benefit of the end user.