This paper presents a system overview of the UAV forcedudlanding site selection system and the results to date. Theudforced landing problem is a new field of research for UAVsudand this paper will show the machine vision approach takenudto address this problem. The results are based on aerialudimagery collected from a series of flight trials in a Cessnaud172.udThe aim of this research is to locate candidate landing sitesudfor UAV forced landings, from aerial imagery. Output imageudframes highlight the algorithm’s selected safe landingudlocations. The algorithms for the problem use imageudprocessing techniques and neural networks for theudclassification problem.udThe system is capable of locating areas that are large enoughudto land in and that are free of obstacles 92.3% ± 2% (95%udconfidence) of the time. These areas identified are thenudfurther classified as to their surface type to a classificationudaccuracy of 90% ± 3% (98% confidence).udIt should be noted that although the system is being designedudprimarily for the forced landing problem for UAVs, theudresearch can also be applied to forced landings or gliderudapplications for piloted aircraft.
展开▼