基于半监督哈希算法的交叉口交通状态识别  

Traffic State Identification of Intersection Based on Semi-supervised Hash Algorithm

作　　者：张立立[1] 王力[1] 赵琦[1] 张玲玉[1] ZHANG Li-li;WANG Li;ZHAO Qi;ZHANG Ling-yu(Beijing Key Lab of Urban Intelligent Traffic Control Technology,North China University of Technology,Beijing 100144,China)

机构地区：[1]北方工业大学城市道路交通智能控制技术北京重点实验室

出　　处：《交通运输系统工程与信息》2020年第1期75-82,共8页Journal of Transportation Systems Engineering and Information Technology

基　　金：北京市长城学者培养计划(CIT&TCD20190304);科技创新服务能力建设-高精尖学科建设(市级)(PXM2019_014212_000020);北京市教委基础科研计划项目(110052971803/013)~~

摘　　要：准确辨识交叉口交通状态是实施有效交通控制策略的前提.传统交通状态识别方法是利用占有率、排队等统计数据设计指标实现状态识别,存在只能从单一角度刻画交叉口交通需求的问题.对此,提出基于半监督哈希算法的交叉口交通状态识别方法.从原始数据丰富特征入手,构建交叉口有效检测区域的图像化模型;将交叉口交通状态识别转化为图像搜索问题,利用监督哈希算法实现基于部分标签信息的图像搜索,进而得到交叉口的交通状态;最后,利用仿真对该方法进行了验证.结果表明,所提方法在识别精度和速度上具有可行性和有效性.Accurate identification of traffic conditions at intersections is a prerequisite for implementing effective traffic control strategies.Traditional traffic state identification method,the state identification is realized by using statistical data design indicators such as occupancy rate and queuing.The traffic state can only describe the traffic demand of the intersection from a single angle.This paper proposes a traffic state recognition method based on semi-supervised hash algorithm.Firstly,starting from the rich features of the original data,the image model of the effective detection area of the intersection is constructed.Secondly,the traffic state recognition of the intersection is transformed into the image search problem,and the supervised hash algorithm is used to realize the image search based on the partial label information.The traffic state of the intersection is obtained.Finally,the method is verified by simulation.The results show that the proposed method is feasible and effective in recognition accuracy and speed.

关 键 词：智能交通 交通状态识别 交叉口图像化 半监督哈希 

分 类 号：U491.3[交通运输工程—交通运输规划与管理]