Robust multi-kernel tracking algorithm for FLIR target

A multi-kernel tracking algorithm based on projection histogram matching and topology constraint was proposed in the mean shift framework, in order to overcome problems caused by partial occlusions and location deviation in the forward looking infrared(FLIR) target tracking. Firstly, target location was predicted by matching the spatio-intensity 2D projection histograms of consecutive frames. Secondly, multiple independent kernel-based trackers were built utilizing the target foreground and background information, which could improve the discrimination of the target model. Finally, topology constraint was used to optimize multiple tracking results, and those trackers with better performance were chosen to generate the final tracking result. Experimental results demonstrate that the proposed algorithm can track target accurately and robustly, and especially in the cases of partial occlusions or location deviation, the average tracking error of the proposed algorithm is much less than that of the traditional algorithm. Therefore, the algorithm has high practicality and feasibility in the FLIR imaging guidance.