基于隐马尔可夫模型的有向地图匹配算法研究

Research of directed map matching based on Hidden Markov Model

地图匹配是指将GPS轨迹映射到真正路网上,获取实际道路上位置的过程。然而,传统的地图算法在处理低频采样数据时(例如,每1~2 min一个采样点)仍然面临着巨大的挑战,此外,这些算法通常是在简单的路网下进行的,并没有考虑道路的双向交通网络。针对这些问题,提出了一种基于隐马尔可夫模型的有向地图匹配算法(DHMM),该算法充分考虑GPS轨迹与道路的相关性以及相邻GPS数据点间的几何特征。结合福州地区真实的出租车轨迹数据,将DHMM算法与点到线(P-L)算法和HMM算法进行比较。实验结果表明,DHMM地图匹配算法在低频和复杂的路网下(复杂的路网由双向道路组成,考虑了道路的方向)匹配准确率均优于P-L、HMM算法。

The process of mapping the GPS trajectories to the real road map and estimating the actual road positions is called map matching.The low-frequency sampling trajectories（ e. g,one point every 1 ～ 2 min） have a great challenge to the traditional map matching methods. In addition,these methods usually perform in the simple road network that does not consider two-way traffic network. In view of these problems,a directed map matching algorithm based on the Hidden Markov Model（ DHMM） is proposed. The algorithm considers not only the correlation between GPS trajectories and roads but also the geometric features between adjacent GPS points. Then it is compared with point-to-line（ PL）algorithm and HMM algorithm based the real taxi GPS trajectories of Fuzhou. The result shows that DHMM map matching algorithm outperforms the above methods（ PL,HMM） in low-frequency under complex road network,which considers the direction of the road,and is composed of two-way traffic.