Box Particle Probability Hypothesis Density Filter for Multi-target Visual Tracking

Box particle probability hypothesis density(BP-PHD)filter for multi-target visual tracking is proposed to improve the computation efficiency compared to the Sequential Monte Carlo probability hypothesis density(SMC-PHD)filter.A fast moving target detection algorithm is given firstly to get the target’s centroid as the measurement by using threshold automatically select frame difference method.Then after the prediction and updating of the box particle PHD filter,the possible deviation of the detection results are corrected,so the target tracking and estimation of targets’number can be realized.Finally,the algorithm of track recognition is designed through the color features and texture features as similarity measure.the algorithm of track recognition is able to compensate the insufficiency of PHD filter.It can distinguish the target track by using the characteristics of the target,and further eliminate the clutter in the target state sets which ensure the tracking accuracy.BP-PHD filter can solve the problems of uncertain measurements and decrease the computational cost.Experiments show that the proposed algorithm can achieve good performance in multi-target video tracking whenever the targets appear,disappear,merge or split,and distinguish the track of different target in real time.