Background Modeling Method Based on Sequential Kernel Density Approximation

Abstract

Background subtraction is a popular moving object detection technique, but its performance is dependent of the accuracy of background model. In this paper, the theory of sequential kernel density approximation is first introduced to background modeling. To this end, a novel background subtraction method for moving object detection is proposed. Various real video sequences have been used to test this method, and comparisons with other standard background subtraction methods also demonstrate that the sequential kernel density approximation is well-suited for background modeling, and the proposed method is effectiveness, it can be efficiently used in various real-time moving object detection systems.