Foreground detection algorithms: A survey

ABSTRACT The task of identifying foreground from a picture or a video is a big challenge in many computer-vision applications. Foreground detection is useful in the security surveillance systems especially in the crowded areas such as parking lots, train stations, airports, movie theaters, hospitals and other mass gatherings. The basic concept to model any background consists of capturing a background image which does not have any moving object in it. However, in many cases, the background does not remain stationary, it may change in the form of illumination changes, tree leaves waving etc. Thus, it is essential to acquire a background model which is not only robust but is also adaptive to background changes. The objective of this paper is to provide a survey of various approaches found in the foreground detection field. The paper concludes by highlighting the advantages and the limitations of these approaches. Keywords: Foreground detection, Background subtraction, Background modeling, Mixture of Gaussians.