Kernel based visual tracking with reasoning about adaptive distribution image

Abstract

A template updating reasoning engine which can deal with fundamental constraints on the spatial-temporal continuity of target's motion is proposed. By analyzing target's continuously adaptive distributions image, a voting method can estimate the tracking window's scale. In updating phase, by making further computation of likelihood of target model and candidate model, both the model and scale can be automatically updated in time. The tracking ability of KBT can be improved. © 2011 Springer-Verlag.