Kernel Based Visual Tracking with Scale-Free Network Model

A novel tracking algorithm is proposed based on kernel based object tracking and scale-free network model. The proposed method regards target model as degree distribution of scale-free complex network model. By taking advantage of scale-free network's characteristic, target's special pixels which have much larger effects than other pixels can be used in tracking process. Based on these special points, target's network can be produced and target scale and target model can also be updated. Experiments show that the proposed algorithm can keep tracking target under conditions of clutter background, varying scales, and partial occlusion. This is the first research about how to use complex network model into the visual tracking algorithm.