基于视频跟踪方法的行人过街状态表达与分析

提出了一种基于视频跟踪的行人过街状态表达与分析方法。采用自适应的混合高斯背景模型实现复杂交通场景下的背景抽取，提取运动行人目标；通过扩展的Kalman滤波算法对过街行人进行跟踪，获取行人的运动轨迹；根据有限状态自动机理论将行人过街过程表达为连续的微观行为状态；在行人过街运动参数的基础上，对行人的过街状态进行推理并分析，得出合理的过街行为方式。通过实际无信号交叉口行人过街视频跟踪实验表明，该方法具有较好的应用性。

Representation and Analysis of Pedestrian Crossing States Based on Video Tracking

Based on video tracking technology,a novel method was proposed to study representation and analysis of pedestrian crossing states.Firstly,a self-adaptive mixture Gaussian model was utilized to extract background for realizing moving pedestrian detection in complex traffic scene.Secondly,the crossing pedestrians' trajectories were obtained through extended Kalman filtering algorithm.Then,the crossing process was represented as a set of micro-states by finite state machine.Finally,according to moving paramete...