A Hidden Markov Model method for traffic incident detection using multiple features

Traffic management is a serious issue in the intelligent transportation systems (ITS). One of the most significant current discussions is traffic incident detection. We have developed an algorithm, referred to vehicle detection based on level set theory and background subtraction, accurate contour of moving object is obtained. The Kalman filtering is applied to predict the possible trajectories of moving object. On this basis, we propose a novel traffic incident detection method based on multiple features and Hidden Markov Model (HMM) classifier. For each pair of vehicles that ever appear together, we extract change of velocity of each vehicle and interaction feature as multiple features. Finally, Continuous density HMM was used for classification of car cash, overtaking two situations. The experimental result showed that the method proposed has good robustness and high recognition rate.