Two-sensor track association based on probabilistic relaxation labeling

Using multi-sensor to detect and track targets has been always a research focus in recent years. And the track-to-track association is an important part of the distributed multi-sensor tracking system. Presently, most of the association algorithms need to do the temporal alignment at the first step, which may bring more errors and degrade the algorithms´ performance due to the impact of noise or inaccurate motion model. Therefore, a novel track-to-track association algorithm between two sensors is proposed in this paper. It synthetically takes account of the target statements and the weighted statistical distances to define the compatibility coefficients of the probabilistic relaxation algorithm. By calculating the compatibility of track pairs it avoids directly measuring the similarities of the tracks from different sensors and implements the association of tracks without temporal alignment. Simulation results show that the proposed method performs better than some classical track association algorithms in many scenarios such as parallel and crossed motion.