基于视频采集数据的跟车模型标定与验证

Calibration and Validation of Car-following Model with Video-capture Data

针对微观交通仿真中跟车模型标定与验证的数据获取精度和数据处理问题,采用视频方法采集了交叉口进口引道上车辆的微观运动录像,利用视频处理软件获取车辆微观运动参数。根据采集数据的特点,采用广义最小二乘法标定模型。选择车头间距作为目标函数的性能指标,对跟车模型进行了标定和验证,误差指标为5.4%,说明标定的模型能很好的反映不同状态下的跟车行为。

For the accuracy of data acquisition and data processing issues of calibration and validation of car-following model in microscopic traffic simulation, video was used to capture micro-movement parameters of vehicles in approaches of intersection. According to the characteristics of data collection, generalized least squares method was used to calibrate model. Headway was selected to be performance indicator of objective function and calibrate and validate car-following. 5.4% error can reflect car-following behavior under the different states.