An improved non-parametric background model and two-level classifier for traffic information recognition

Acquirement of real-time and overall traffic information is very important for improving road network efficiency and reducing traffic congestion. This paper proposed an improved non-parametric background model to segment the moving vehicles from traffic videos with limited computational complexity and space complexity. With the analysis of characteristics of traffic parameters, a two-level classifier is proposed for automatic recognition of traffic information. The results from automatic recognition have high coincidence rate with those from expert classification.