Learning and Classification of Maneuver Behaviors Based on Deep Belief Networks

Abstract

Cognition of air target maneuver behavior is an important basis for identifying air combat intentions. In order to enhance the airborne radar’s perception of air combat situation, this study developed a framework for activity classification of observed typical maneuver behaviors of air targets. A target athletic behavior learning matrix was constructed to make regularized description of multiple types of behaviors under multiple classification goals; also, a group of network parameters was determined, which was suitable for motion behavior classification. Finally, experiments were used to verify the effectiveness of the proposed method.