Visual Tracking Model Based on Feature-Imagination and Its Application

It is an important and challenging problem to tack variant-pose target, especially when the appearance of target changes during the period of occlusion. The traditional feature matching methods can not work well under the situations of target pose varying. This paper analyzes the characteristics of human visual tracking, and proposes a visual tracking model based on feature-imagination. Human tracking is realized by using this model combined with the traditional histogram matching method. The experiment results show that our method can perform feature-imagination from the features of human front to the features of human back. Compared with the traditional histogram matching method, the proposed method has obvious advantages in tracking variant-pose target.