Research of particle filter tracking algorithm based on structural similarity

In this paper, we address the problem of tracking a randomly moving target in the similar background. Traditional color particle filter algorithm cannot actively judge the interference created by similar background, and cause that particle filter estimate by mistakes. So we propose and implement the tracking by using structural similarity as a standard for judging tracking states. We establish unified color scale for background to extract color histogram so that we can real-time track moving target. Further we calculate the average structural similarity between the target and background, which is weighed by particles weight, to establish tracking model. The performance of the considered algorithm which is illustrated through computer simulations shows the improved color particle filter that is more effective when both the moving target and the background are very similar.