Adaptive Feature Selection for Infrared Object Tracking

This paper proposes a novel object tracking algorithm for the infrared object tracking. To improve the discrimination ability of the object model, the gray features which can be used to distinguish the object from its surrounding background is chosen to represent the object. This algorithm defines a discrimination function of gray features, and the "good" gray features are selected by analyzing the discrimination function in a large scale. To deal with the variation of the intensities of the infrared object and the background during tracking, an adaptive feature updating strategy is proposed. The two dimension histogram is built based on the selected features to model the object, and the object tracking is achieved by mean shift algorithm. The experimental results using real-life infrared image sequences are shown to validate the robustness and efficiency of the proposed algorithm.