Robust infrared target tracking with Kalman prediction sampling and spatiogram representation

Target states sampling strategies and observation probabilistic models are main factors on the infrared target tracking performance under the theory framework of particle filters.In order to improve the performance of the infrared target tracking,a novel method was proposed,which was based on Kalman prediction sampling and spatiogram target-representation.Kalman prediction sampling,which could combine target observations into the importance proposal distribution by the Kalman prediction process,was adopted to implement infrared target state sampling for particle filtering.Robust infrared targets were represented by the spatiogram,which could capture the spatial information of infrared targets.With the spatiogram target-representation,observation probability models were constructed by computing the Bhattacharyya distance between the reference target's spatiogram and target samples' spatiogram.Three infrared target tracking experimental results demonstrate the method is effective and robust for different scenes,such as the sensor ego-motion scene,the unstable radiation scene,and the sea-clutter scene.