Abstract: Magnetic vehicle detector is a rising traffic flow data collection technology in recent years. In the related research field, vehicle speed estimation based on single sensor is one of the hot spots. This paper introduced the magnetic vehicle detection technology. The distribution statistics of vehicle length on urban road network was analyzed. Under certain reasonable assumptions, the vehicle class composition identification based mean speed estimation algorithm was then put forward. In the algorithm, the OTSU method was used to classify vehicles into small and large vehicles. On urban road network, small vehicles appeared mostly and the vehicle lengths distribution was centralized. According to the statistical characteristics, mean vehicle speed was calculated based on only small vehicles data in the algorithm. Finally, field experiment was conducted on road section in Beijing and the algorithm was verified on the Matlab platform. It was concluded that, the algorithm was with high accuracy and stability. The accuracy of calculated mean vehicle speed exceeded 85%.

Key words: intelligent transportation systems, magnetic sensor, vehicle speed estimation, vehicle class composition identification, OTSU method

摘要： 地磁车辆检测器是近年来兴起的新型交通流信息检测技术，而利用单节点检测数据完成车速估计是该技术研究的重点之一. 本文首先对地磁车辆检测技术予以简要介绍，然后基于对道路车辆车长分布规律的分析，在合理假设的前提下，提出了车流特征分区的车速估计算法. 该算法充分考虑实际应用中各类车型组成情况，应用最大类间方差法实现大型车与小型车的分类，基于城市道路小型车居多且车长分布集中的特点，仅利用小型车通过时间检测数据完成对平均车速的估计. 最后，选取实际路段开展试验，并基于Matlab平台对该算法进行了验证. 结果表明，本文提出的车流特征分区算法稳定性好，平均准确率达到85％以上.

关键词: 智能交通系统, 地磁传感器, 车速估计, 车流特征分区, 最大类间方差法