Aircraft tracking plays a key and important rolein the Sense-and-Avoid system of Unmanned Aerial Vehicles(UAVs). This paper presents a novel robust visual trackingalgorithm for UAVs in the midair to track an arbitrary aircraftat real-time frame rates, together with a unique evaluationsystem. This visual algorithm mainly consists of adaptivediscriminative visual tracking method, Multiple-Instance (MI)learning approach, Multiple-Classifier (MC) voting mechanismand Multiple-Resolution (MR) representation strategy, that iscalled Adaptive M3tracker, i.e. AM3. In this tracker, theimportance of test sample has been integrated to improvethe tracking stability, accuracy and real-time performances.The experimental results show that this algorithm is morerobust, efficient and accurate against the existing state-of-arttrackers, overcoming the problems generated by the challengingsituations such as obvious appearance change, variant sur-rounding illumination, partial aircraft occlusion, blur motion,rapid pose variation and onboard mechanical vibration, lowcomputation capacity and delayed information communicationbetween UAVs and Ground Station (GS). To our best knowledge,this is the first work to present this tracker for solving onlinelearning and tracking freewill aircraft/intruder in the UAVs.
展开▼