Auto-selection of optimal visual representation for a flying robot based on altitude control

Recently, multi-robot cooperation with the grand and aerial robot (UAV) has been attracting attention for the mission of navigation, mapping, surveillance, and target search. Autonomous UAV requires sound visual recognition system which has to cope with limited resolution to identify the ground targets and frequent disturbances by the wind and lighting conditions. This paper propose a novel approach to enhance a capability of visual perception of UAV in the flight, where the UAV autonomously evaluates the type of representation feature to obtain most salient recognition available in the current condition. In the case that UAV has to use a specific type of visual feature, the altitude control makes it possible to search the best observation point for the object recognition with the current visual feature. The experiment results imply the feasibility of the proposed approach.