REAL-TIME ILLUMINATION COMPENSATION ON DYNAMIC BACKGROUND FOR CROWD ANALYTIC SURVEILLANCE

Background modeling is one of the key steps in any visual surveillance system. A
good background modeling algorithm should be able to detect objects/targets under
any environmental condition. The influence of illumination variance has been a major
challenge in many background modeling algorithms. Previously developed algorithms
which intend to solve this issue produced poor object segmentation or consumed
substantial amount of computational time, which made them not implementable at
real time. In this thesis we propose a novel background mod:elin0 ITtdhod b<1:>ed on
Phase texture features. The proposed method uses Phase texture features to overcome
the effect of illumination variance, while preserving efficient background/foreground
segmentation and computational efficiency. The proposed method is been validated
qualitatively and quantitatively using various crowd analytic databases. Moreover,
the results of our method is been used with other behavior learning algorithms to
improve their efficiency.