基于车牌时空数据的城市热点交通线路挖掘

Mining urban hot routes based on spatio-temporal license plate number data

交通摄像头在智能交通系统(ITS)中的作用日益重要,其主要功能为车牌识别。本文提出了一种从车牌时空数据中挖掘城市热点交通线路的方法。其中,车牌时空数据由部署在城市不同道路的交通摄像头不断进行车牌识别得到。实现该目标存在以下挑战:首先,一辆车的轨迹(由车牌时序数据代表)通常只占城市热点交通线路的一部分。其次,车牌识别存在高度的不确定性(如遗漏和错误),使得现有模式挖掘算法难以发现完整的城市热点交通线路。针对以上问题,本文提出了由2部分构成的方法。首先,该方法提出了一个基于子模式拼接的挖掘算法,从车牌时空数据中挖据出候选城市热点交通线路。然后,该方法基于一个聚类排序算法从候选城市热点交通线路中挑选出代表性城市热点交通线路。本文基于真实车牌时空数据对提出的方法进行了评测。

Traffic camera plays an important role in intelligent transportation systems(ITS).A major function of traffic camera is license plate number recognition.This paper focuses on discovering city-wide hot routes using license plate number data recorded by traffic cameras deployed throughout the city.This task is challenging due to the following two reasons:First,a vehicle trajectory could usually contribute to only a small portion of a hot route.Second,the high degree of uncertainty of license plate number data makes the existing mining algorithms ineffective.Aiming at these problems,a two-phase method is proposed.First,it extracts hot routes by aggregating the license plate number data from multiple traffic cameras and vehicles.Second,it compresses the mined hot routes based on a clustering and ranking algorithm.The proposed method is evaluated based on real-world license plate number data from a city-wide traffic camera system.