Disorder detection approach to background modelling for traffic surveillance system

Abstract

This research proposes a new background subtraction technique with disorder detection approach for traffic surveillance system. By applying entropy function, we recognize the disordered frames (DF) from image sequence which adversely affect to background images, then remove them from background modeling step by using 3-state process. The background and foreground are produced by Gaussian mixture model from the other qualified images. As the result, the new approach obtains astonishing result under real-life traffic condition. It also decreases computation time for background subtraction step in real-time system. Keywords: background subtraction, background modeling, disorder detection, mixture of Gaussian model, entropy.