An image feature descriptors-based recovery activation metric for FLIR target tracking

This paper describes a novel method to detect false alarms in target tracking applications based on image-based feature descriptors. The number of matches between the local feature descriptors in the reference target and in the candidate target is used to assess the reliability of the tracking algorithm used and, in case, to activate a suitable recovery mechanism. The proposed approach has been evaluated on a series of Forward Looking Infrared (FLIR) sequences provided by the U.S. Army Aviation and Missile Command (AMCOM).