交通冲突判别的车辆运动模式分析方法

Vehicle Motion Pattern Analysis Method for Traffic Conflict Discrimination

摘要:

综合运用交通冲突理论与视频图像处理技术，建立了一种基于车辆运动模式的交通冲突判别方法.根据安全金字塔原理与车辆运动模式，提出全面可量化的交通冲突严重性判别指标——碰撞可能性指标及其计算公式.针对轨迹样本的时空转移特点,确定各模式类隐马尔可夫模型结构，建立了基于隐马尔可夫模型的车辆运动模式学习与预测算法.最后，建立了完整的视频分析系统并在实际交通场景中进行了测试，结果验证了本文方法的可行性与适用性.

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.