Research on extended target tracking with partial region feature

Abstract

Since extended targets have some characteristics that are different from generic objects, most traditional target tracking algorithms cannot be used anymore. Moreover, image blurring, illumination change, rotation, and occlusion usually exist during extended target tracking, which brings huge challenges for feature extraction and tracking. In this paper, a tracking algorithm based on partial region feature is proposed. Firstly, a method for partial feature region selection based on entropy image is put forward. And then, by using invariant moments correlation matching, a partial region feature tracking strategy is analyzed, in which the matching and template updating methods based on paraboloid fitting are mainly discussed. A simulation results shows that it has a good performance for extend target tracking.