Taeget Tracking Algorithm Based On Particle Filter

Computer video among a very important part is the video tracking, and now the video tracking which has been widely used in national defense, science and technology, aviation and national life. The more representative video tracking is human-computer exchange, security monitoring, Anti-aircraft missiles and traffic management.Video tracking can track the target effectively by particle filter algorithm.The main steps of video in the video sequence is the track target doing target detection, extraction, identification and tracking. So you know the tracking target state, which include its speed, trajectory, position and acceleration information, thereby further information to predict the next target tracking and tracing, and finally achieve the goal of understanding the behavior tracking. The main factor of resulting in the tracking failure of video tracking is the more complicated background, tracking barrier, the changing appearance and the changing scale of tracking target.Using conventional particle filter in the video tracking process to track the similar color and long time completely block is less than ideal.So,first,for the issue of similar target block in the video tracking process,tracking the similar type target by using the multi-observe model and combined with adaptive reference model.Second, multi-observe model includes color observation model and observation model of direction and gradient.Finally,against calculation method of the color observed model similarity, this paper presents an improved HSV color space calculation method to calculate the similarity.For a long-time blocked completely issue in the video tracking,the article in the particle filter framework,judging whether is blocked completely by block matching stencil.If the judgment is completely blocked then stop updating the reference model, and finally the use of a cyclical global search mode for searching and tracking targets again.The experiments show that in the video track encountered similar object occlusion and complete occlusion,using improved algorithm is more effective in tracking effect in timeliness and robustness than particle filter algorithm based on RGB model.