Abstract: The current intelligent transportation systems can only be to monitor the traffic flow characteristics, unable to segment different states of the traffic. On the basis of fully study of traffic monitoring principle, put forward a kind of more traffic state recognition based on intelligent visual identification system, using the monitoring image frame differential variance interval, update the image background, introduction of pixels change control rules as core software algorithm, combining with traffic condition monitoring data, detect jams, pedestrians and traffic event parking state, etc. In real system testing shows that system can timely and accurately detect the relevant traffic incident, and has high detection precision..