Vision-based navigation and exploration strategies for unmanned helicopters in disaster scenarios

Abstract

This paper presents two vision-based aspects to solve problems for unmanned aircraft within disaster scenarios. Such scenarios are often related to low-altitude flights in urban or mountainous areas where satellite navigation can be disturbed. To cope for possible signal dropouts and to enable flights with low-cost inertial systems, visual navigation strategies are developed. The paper presents a method to get 3D movements from a stereo camera onboard an unmanned helicopter, evaluates the image processing algorithms from flight test data, and concludes that this visual navigation strategy is suitable as an additional sensor for flight state estimation. The other aspect tackled in this paper is the localization of target coordinates within exploration missions. The presented algorithm is based on known strategies in remote sensing. It is optimized here for real-time use onboard remotely operated systems. It is shown that targets can be localized with only few meters of absolute error without any external hints, with satellite navigation as the most relevant error source. As a result, supporting applications like the localization of victims are feasible with the presented strategy.