A new tracking method based on Mean-SIFT and particle filter

In order to solve the multi-target tracking problems in video sequence, this paper presents a algorithm integration Mean-Shift(MS) and particle filter(PF) called KMSPPF to tracking multi-target. The algorithm uses the K-means clustering results as the optimal input to the Particle Filter, Mean Shift follows by resampling and then particles converge to the true state of the target, thus overcomes the traditional particle filter degradation and lessen the time of computing; it can also solve the problem of target occlusion. The experimental results show that the algorithm can reduce the computational cost while tracking multi-target, and ensure the performance simultaneously.