The appearance model for visual tracking by using twins-updating template

The twins-template algorithm is proposed in this paper to track object in videos. First, the initial appearance model of affine transformed in the first frame is specified as a static template and the variable model is the last updating appearance model specified in last several frames, which is different from traditional particle filter. The current location of object is decided by the twins-template collectively. Second, a neighbors window including the appearance model in every frame is processed by using histogram equalization for reducing the illumination variance of background. At last, the dynamic target updating scheme keeps track of the most representative particles throughout the tracking procedure. The proposed approach shows excellent performance on several image sequences. © 2012 by Binary Information Press.