Summer UAV Search and Rescue Field Trial

May 29, 2008

Michael A. Goodrich

      Using the Virtual Cockpit user interface, developed by the BYU MAGICC Lab and Procerus, we were able to successfully guide the UAV through a search path and localize the mannequin. Set-up time was quick, and the UAV was flying after twenty minutes or so at the site. After a little more than an hour of searching the UAV was brought in again for a battery change and a lunch break. The UAV was launched again for a second search, in which it found the dummy after only fifteen minutes or so of searching. Delays due to technical difficulties were minimal. Adjustments to the UAV were simple and quick, the hardware was dependable, and the take-off, navigation, and landing went smoothly.

      The Mosaic software written up by Bryan Morse's Computer Vision lab worked very well (although missing a few needed features) and on the whole the images produced by Mosaic were much better for searching the ground than the straight video feed. As for the hardware, some adjustment was needed on camera (the gimbled camera gave a much steadier image, but the colors were greatly white-washed out). On the whole the UAV flew without trouble, but in one instance its battery did give out just as it was coming in for the landing.

      Definitely an area in need of improvement was the fact that without time stamps, coordinates, or pausing feature on the video feed, even after searchers caught sight of an object they wanted to investigate, it took far too much time to (a) find the object on-screen again and (b) direct the ground team to where it was. Multiple searches with the UAV over the same spot were required, and there were quite a few problems trying to figure out where the UAV was when it took the video in which the 'missing person' could be seen. Research into integrating map information with terrain information, video, and video mosaicking is ongoing.