Modeling movement and probability of area: Although this is a research area dear to my heart, I think that it will probably only have impact in the long term. Three things must be addressed before it can have high impact. First, UAV's must be perceived as a useful tool by the search and rescue community. Second, we've got to get the models right so that they really help rather than just provide pretty pictures. Third, they have to complement emerging Geographic Information Systems intended to support the incident commander or the ground searcher.
Providing imagery in adverse weather or in extremely rugged terrain: If a search is a multi-period search, then it will be more likely that manned aircraft will be used to help the search. Unless we approach the same reliability as the manned aerial searches at a much cheaper price, it seems unlikely that the UAV technology will be used. Instead, we might want to find ways to provide aerial imagery in situations where it is impossible or dangerous to use manned aircraft. Better communications and more robust flight path planning would be essential in such circumstances.
Providing IR imagery at night: Many searches stop or are dramatically scaled back at night. If we could fly reliably enough, then we could gather IR imagery at night that, when combined with visible-spectrum imagery from the daytime, could help focus search in the right areas (or at least minimize the commitment of ground searchers to unlikely regions).
Enhancing probability of detection using image enhancement techniques: In challenging search areas, it would be very useful to enhance imagery to make man-made objects stand-out against the background. Since false alarms and missed detections dramatically affect the potential benefit of such technology, it is important that image enhancement be carefully studied.
New idea 6/4/09 Automatic identification of trail-based probability of areas:“ Assigning probabilities to the likely location of a missing person is a challenging task. A relatively new approach to assigning such probabilities is called “trail-based probability” ://www.sarinfo.bc.ca/Trailpoa.htm. This technique assumes that the missing person intends to follow a trail or other linear feature, but departs from this linear feature at some decision point. Decision points are supposed to be identified by ground searchers traversing the trail. Would it be possible to do this more quickly by having the UAV follow the trail, gathering imagery as it spirals over the waypoints delineating the trail, and automatically identifying possible decision points using computer vision techniques. These decision points could then be served as points of importance into a probabilistic model for how the missing person moves through the terrain.
Back to top