In this article, a novel pixel based object detection framework is proposed that leverages dual type pixel-level information to construct the background model.

Video surveillance is of fundamental and critical importance for analysis and understanding interesting objects. An improved sampled-based model in this paper is introduced to detect these objects.