Probability Based Approach for Target Detection and Pose Estimation in Infrared Images

This paper presents the challenging issue of target detecting and pose estimation in Forward Looking Infrared (FLIR) sequences. Target detecting and pose estimation are formulated as one process. Specifically, a new mixture of Gaussians model is proposed to detect target appearances. A probability based method is used to estimate the target position and size, where the detected appearance plays an important role. Our probability-based algorithm is compared against the existing methods and experiments on the FLIR dataset validate its effectiveness.