基于亚线性MG替换策略-网络流的动态车流量检测方式  

A DYNAMIC TRAFFIC FLOW DETECTION MODE BASED ON SUBLINEAR MG REPLACEMENT STRATEGY AND NETWORK FLOW

作　　者：沈智勇 苏翀 沈智威 孙厚权 周扬[1] Shen Zhiyong;Su Chong;Shen Zhiwei;Sun Houquan;Zhou Yang(Jiangsu University of Science and Technology,Zhangjiagang 215600,Jiangsu,China;Soochow University,Suzhou 215000,Jiangsu,China)

机构地区：[1]江苏科技大学,江苏张家港215600 [2]苏州大学,江苏苏州215000

出　　处：《计算机应用与软件》2018年第9期32-37,共6页Computer Applications and Software

基　　金：中国博士后科学基金项目(2016M600430);湖北省水电工程智能视觉监测重点实验室开放基金项目(2016KLA08)

摘　　要：由于道路复杂、检测器数量基数大,往往采用随机化方法来估算道路车辆动态流量。基于图像处理的方式对静态车辆有很好的监测效果,对动态车流量检测效果一般;纯红外的方式很难对数据进行利用且在数据量大的情况下有很大的限制。基于红外收集点数据,采用基数桶-块状链表结构存放数据以限制存储空间;采用改进的Misra Gries(MG)替换算法逐步修正偏差量;利用大数定理计算各段路径所需费用并构建网络图,从而求出该图中的最大流量并判断道路交通情况。用该方法对红外数据进行分析处理,对于高速的动态车辆具有很好的效果,数据有效性大于90%。实验结果表明,基数桶结构能很好地适应Misra Gries替换算法,对于纯红外检测的算法能提高近百倍的计算量。Due to the complexity of roads and the large amount of detectors, a randomized method is often used to estimate the dynamic flow of toad vehicles. The way based on image processing has a good detection effect on static vehicles and has a general effect on dynamic traffic flow detection. It is hard for pure infrared to use data and there are great limitations in the condition of large amount of data. The paper used the cardinal number bucket-block list structure to store data so as to tstrict storage space. The improved Misra-Gries （MG） replacement algorithm was adopted to gradually correct the deviation. We utilized the law of large numbers to calculate the cost of each path and build the network diagram in order to find the maximum flow in the graph and judge the traffic condition of the toad. The method was used to analyze and process infrared data. It has a good effect in high speed dynamic vehicles and the data validity is more than 90%. The experimental results show that the cardinal number bucket structure can well adapt to MG algorithm and increase the computation amount by nearly 100 times for the pure infrared detection algorithm.

关 键 词：随机化 基数桶 Misra Gries算法 网络流 车流量检测 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]