面向机动车出行OD监测的目标路段选择算法

机动车在感兴趣区域（例如景点或小区）之间的出行起始和结束（Origin-Destination,OD）信息反映了居民使用机动车出行的活动需求分布,是建立城市智能交通系统中交通需求分析与管理的基础信息之一。数量充足的监测设备能够收集的机动车出行信息更为精细,最终构建的OD信息可用性更强。然而在实际应用中,设备数量往往因为预算情况而存在限制。在设备数量有限的情况下,分析如何选择监测路段来实现OD信息可用性的最大化具有重要应用价值。考虑到感兴趣区域的空间精细程度直接影响监测设备的需求,首先采用层次聚类思想调整感兴趣区域的空间精细程度;然后,根据道路交叉口车流量守恒原则,探测冗余监测路段来进一步降低对监测设备的需求。本研究基于上述2步操作来实现给定数量设备监测能力的最大化。该算法以摄像头为例,在深圳市大鹏半岛区域进行了实验。结果显示,该算法能够支持在不同摄像头数量限制的情况下制定监测路段的选择方案,表明了算法的有效性。

A road section selection algorithm for monitoring the OD flow of motor vehicle travels

The origin-destination (OD) information of motor vehicles serves an important foundation in urban traffic analysis and intelligent transport system management. Currently, most of the location-allocation algorithms of traffic detectors in transportation system focus on detecting traffic conditions (e.g., travel speed) at major junctions of road networks. However these algorithms fail to completely monitor the OD information of motor vehicle travels. This study proposes an algorithm to select the road sections where traffic detectors should be installed for the purpose of monitoring the OD information of motor vehicles traveling between regions of interest (ROIs) (e.g., residential communities and shopping malls). Two methods are adopted in this algorithm to maximize the usability of the detectors. First, since the demand of the detectors relies on the number of the road sections connected to the ROIs and the spatial resolutions of the ROIs affect the volume of connected road sections, the spatial resolutions of the ROIs are adjusted according to the closeness between the ROIs by a hierarchical clustering algorithm. During this process, special ROIs, for which the ODs need to be monitored independently, can be set to avoid being merged with other ROIs. Second, the redundantly monitored road sections are detected based on the conservation law of the traffic flows at a crossroad. This algorithm was examined in the area of Dapeng, Shenzhen. Specifically, we first used a museum and a road entrance to test the effectiveness of the special ROIs setting. Then we compared the outcomes of the hierarchical clustering algorithm using three different distance measurements between clusters, namely single linkage, complete linkage and average linkage. Third, the effectiveness of the redundant monitor road sections was examined at a crossroad. Last, we tested the effectiveness of the proposed algorithm when the supply of cameras were limited to 10 and 20, respectively, based on the simulation result of ODs between different locations. The results suggest that the proposed algorithm can effectively support the policy-making of selecting the target road sections to monitor the OD information of vehicle travels when the supplies of the detectors are limited under different situations.