Vision based terrain recovery for landing unmanned aerial vehicles

We propose a system for landing unmanned aerial vehicles (UAV), specifically an autonomous rotorcraft, in uncontrolled, arbitrary, terrains. We present plans for and progress on a vision-based system for the recovery of the geometry and material properties of local terrain from a mounted stereo rig for the purposes of finding an optimal landing site. A system is developed which integrates motion estimation from tracked features, and an algorithm for approximate estimation of a dense elevation map in a world coordinate system.