Meanshift tracking based on visal
4.0
Meanshift by Comaniciu first put forward, is an algorithm for tracking non-rigid objects. In order to characterize the target, you first need to select a feature space. Reference objects, represented by the probability density function of q in the feature space, such as reference models you can choose color probability density function. Without loss of generality, it can be considered the center of the target model at the Centre of their space. Next frame candidate targets are defined as in location y, with probability density function p (y) to represent. In order to meet real-time requirements, respectively, using non-parameter histogram in ^q and ^p (y)