Particle Filter Algorithm Based On Motion Information

Visual target tracking is a key problem in computer vision, which has a wide rangeof applications in military guidance, visual surveillance and medical diagnosis, etc. Thetask of visual target tracking is to track the moving target in a sequence of images andthen to provide an important data source for visual analysis and understanding. So it isof great significance in theory and practical applications to study target tracking undercomplicated background.About the visual target tracking, the conventional particle filter uses the statetransition distribution as the proposal distribution, from which the particles are drawn ineach iteration. However, the transition distribution does not take into account the currentobservation, and thus many particles drawn from it will not only consume a lot ofcomputing time, and will reduce the effect of target tracking. In this paper, basedparticle filter, we combine motion vector in H.264video compression video codingstandard to determine the quality of all particles in each stage, replace the particles thataffect the tracking performance, and retain good quality particles to improve theaccuracy of target tracking.Experimental results prove that compared with the traditional condensationalgorithm, proposed improved particle filtering algorithm has better accuracy for visualtarget tracking.