Vehicle Motion Pattern Analysis Method for Traffic Conflict Discrimination

Abstract:

A method of traffic conflict discrimination based on vehicle motion pattern is established using traffic conflict theory and video processing technology. According to the principle of safety pyramid and vehicle motion patterns, this paper presents a comprehensive and quantitative traffic severity indicator which is collision probability index and its calculation formula. In view of the spacetime tranfer characteristics of sample trajectories, the structures of the hidden Markov models are first determined for various trajectory clusters, and then the learning and prediction algorithm of vehicle motion patterns based on hidden Markov theory is established. Finally, the complete video analysis system is realized and tested in the real traffic scene. The experimental results have confirmed the feasibility and reliability of the proposed method.