Scale-Aware Real-Time Scene Reconstruction for a Micro Aerial Vehicle
Principal goal for project:
A study into vision-based SLAM for small aerial vehicles, and a user-interface to control such a vehicle equipped with an onboard camera.
This thesis sets out to produce a 3D reconstruction of the environment explored by a micro aerial vehicle (MAV) equipped with only a single camera and inertial measurement unit. To achieve our goal, we combine a scale-aware variant of the localisation and mapping algorithm PTAM with an incremental real-time reconstruction algorithm. We demonstrate that it is possible to extract the distance and angle between the MAV and a target selected within the reconstruction with reasonable accuracy in the case of a small, simple scene. Additionally, we show a correlation between the amount of data in the map and improved accuracy of distance measurements, such that larger maps should yield better measurements. We posit that this is a proof of concept that our scale-aware reconstruction could be utilised in allowing way point based control of the MAV and more.