Simulation testing is an initial phase for the development of autonomous vehicle system. The developers can ensure the model they use behaves as close to the actual vehicle as possible in this cycle. This paper proposed a novel traffic scene modeling approach using image sequences and road GIS data. Such method was contributed to building a simulated environment for autonomous vehicle software systems testing and validating as well as its various algorithms. Based on the proposed approach, a traffic scene simulator, the RoadView is realized. Compared with the existed robot testing simulators such as the USARSim and the Gazebo, the RoadView can provide a more photorealistic scene. This simulator is employed to test performance of autonomous vehicle and evaluate its self-driving tasks. In such process the proposed approach is proved workable and effective for autonomous vehicle simulation testing.