A Vision Based Forced Landing Site Selection System for an Autonomous UAV

• DOI: 10.1109/issnip.2005.1595612
• Publication Date: 2008-07-18T11:12:17Z
• Authors: D. Fitzgerald; R. Walker; D. Campbell
• Keywords: 629; 0209 industrial biotechnology; Forced Landing; Computer Vision; UAV; 02 engineering and technology; Classification

This paper presents a system overview of the UAV forced landing site selection and results to date. The problem is new field research for UAVs this will show machine vision approach taken address problem. are based on aerial imagery collected from series flight trials in Cessna 172. aim locate candidate sites landings, imagery. Output image frames highlight algorithm's selected safe locations. algorithms use processing techniques neural networks classification capable locating areas that large enough land free obstacles 92.3% ± 2% (95% confidence) time. These identified then further classified as their surface type accuracy 90% 3% (98% confidence). It should be noted although being designed primarily UAVs, can also applied landings or glider applications piloted aircraft.