The Study of Improved Particle Filtering Target Tracking Algorithm Based on Multi-features Fusion

In view of the shortcomings of traditional particle filter which is lacking of utilizing current observational information, this paper proposes a multi-featured fusion tracking algorithm based on simulated annealing to improve particle filter. The proposed method solves the problem of large amount of computation and lack of particle number in high dimensional state. A hierarchical random search annealing method is used to generate a better proposal distribution in the Monte Carlo importance sampling. In the likelihood approximation, this paper integrated image feature attribute of colors and edges to generate weight function in the different annealing layer by weighting. Using this method to track the moving objects with complex background and occlusion, the experimental results show that the proposed method has high tracking accuracy and strong stability.