Feature Selection Method For Single Target Tracking Based On Object Interaction Models

For single-target tracking problem Kernel-based method has been proved to be effective. A tracker which takes advantage of contextual information to incorporate general constraints on the shape and motion of objects will usually perform better when compare to the one that does not exploit this information. This is due to the reason that a tracker designed to give the best average performance in a variety of scenarios can be less accurate for a particular scene than a tracker that is attuned (by exploiting context) to the characteristics of that scene. The use of a particular feature set for tracking will also greatly affect the performance. Generally, the features that best discriminate between multiple objects and, between the object and background are also best for tracking the object.