A Tracking Algorithms Based on Particle Filter Using Adaptive Optimization

For the problem that poor quality of particle and large amount of computation, a particle filter tracking method based on adaptive optimization is presented by using color clue of object. In our method, 24-bit color pixels are used to establish a kernel histogram of object model, and then adaptive models are built according to Bhattacharyya coefficient of object model in previous frame. In addition, mean shift is selected adaptively according to current environment to optimize the particle. Finally, a large number of comparisons experiments show the presented method be effective and self-adaptive.