为研究交叉口绿闪信号期间行人过街行为特征，弥补现有行人仿真模型在此场景下的不足，首先基于无人机高空摄像采集的交叉口交通视频数据，采用Tracker影像分析软件提取行人轨迹与速度信息，对比分析了绿闪与常规绿灯信号下行人过街的速度与避让行为特征。然后，结合实际数据改进传统社会力模型中的目标速度参数，构建了绿闪信号下的行人过街行为模型。最后，借助AnyLogic平台开展仿真实验，将仿真结果与实测行人过街速度进行对比。结果表明：绿闪时段行人平均步行速度为1.92 m/s，显著高于绿灯时期的1.48 m/s，部分行人以2.5~3.3 m/s的速度快速通过；绿闪时段行人在过街过程中普遍表现为“加速-减速-平稳”三阶段速度变化趋势；行人避让行为发生频率在绿闪场景下减少26.4%，避让触发距离平均延长1.22 m；行人避让横向偏移量保持在0.52 m左右，与绿灯时期无显著差异。仿真结果显示，模型速度曲线与实测数据拟合良好，平均模拟精度为95.2%，较已有相似场景研究的精度提高了6%。这表明改进后的模型能够较好地还原绿闪信号下行人过街的速度变化，可为信号交叉口行人仿真提供参考。

To investigate pedestrian crossing behavior characteristics during the flashing green signal phase at intersections and to compensate for the deficiencies of existing pedestrian simulation models in this scenario, firstly, based on the intersection traffic video data collected by high-altitude drone cameras, the Tracker video analysis software was used to extract pedestrian trajectory and speed information. The speed and avoidance characteristics of pedestrians under flashing green and regular green signals were comparatively analyzed. Then, by combining actual data to refine the target speed parameter in the traditional social force model, a pedestrian crossing behavior model under flashing green signals was constructed. Finally, simulation experiments were conducted using the AnyLogic platform, and the simulation results were compared with the measured pedestrian crossing speeds. The results show that the average walking speed during the flashing green phase was 1.92 m/s, significantly higher than 1.48 m/s under regular green signals. Some pedestrians crossed quickly at speeds of 2.5~3.3 m/s. Pedestrians crossing during the flashing green phase commonly showed a three-stage speed variation trend of "acceleration-deceleration-stabilization". The frequency of avoidance behavior decreased by 26.4% under the flashing green condition, and the average avoidance trigger distance was increased by 1.22 m. The lateral offset of pedestrian avoidance remained around 0.52 m, showing no significant difference compared to regular green condition. The simulation results show that the velocity curves generated by the model fit well with the measured data, with an average simulation accuracy of 95.2%, representing a 6% improvement over existing studies under similar conditions. This indicates that the improved model can effectively reproduce pedestrian speed variations under flashing green signals and can serve as a useful reference for pedestrian simulation at signal intersections.