Omnidirectional Vision Tracking with Particle Filter

Visual tracking is one of the most important topics of computer vision. Challenging situations occurs when the time between frames is long enough to show significant spatial jumps of the target being tracked or when the target has unexpected motions. In this paper, we offer an alternative to these situations. We use background subtraction to find the foreground regions (regions that move between frames). Using the foreground regions and a modified version of the well-known particle filter we are able to track targets in an omnidirectional sequence with a low frame rate (one or two frames per second). Our approach can tackle unexpected discontinuities and changes in the direction of the motion