基于交通波检测的交通参数获取研究

发布时间：2018-05-27 21:39

本文选题：计算机视觉 + 交通波　；参考：《北京工业大学》2015年硕士论文

【摘要】：随着城市汽车保有量的持续快速增长,城市道路交通拥堵问题日益严重,迫切要求提出更为合理的交通控制策略,而交通参数的准确获取是优化交通信号控制策略的前提。因此,采用先进信息技术准确获取交通参数对改善城市交通具有重要意义。近年来,基于机器视觉的智能交通参数检测算法应运而生,应用计算机和图像处理技术检测各种交通参数,分析交叉口交通流特征成为研究热点。尤其是在交叉口早晚高峰时段车辆排队严重,采用视频方法如何实现高精度的参数提取具有挑战性。对此,本文针对交叉口处基于交通波检测的交通参数获取进行了研究,主要研究内容包括以下几个方面:1.提出了一种基于人工标定的交通参数真实数据获取方法。该方法利用VIPER智能软件,获取车辆每一时刻在视频图像中的像素坐标。利用摄像机模型,将像素坐标转化为道路平面空间坐标,根据标定车辆的时间-位置确定排队长度及停车延误等交通参数,根据所有标定车辆的起停变化拟合出真实的交通波到停车线距离变化曲线,获取准确的交通波位置信息。2.分析了现有的基于视频的典型交通波检测方法。对基于单摄像机的复式伸缩窗算法、基于对偶像机的决策层数据融合算法及像素层数据融合算法流程进行了分析,并将交叉口处基于人工标定获取的交通波位置数据与三种典型交通波检测算法得到的数据进行对比,通过建立评估指标来分析各算法性能的优劣。3.开展了基于交通波检测获取交通参数的研究,包括排队长度、停车延误、波速等交通参数的提取方法。通过将各参数与停车波和起动波建立数学关系,推导出相应参数的计算方法。其中平均停车延误的计算要根据周期内只存在一次排队和存在两次排队分别讨论。4.以实验为基础,开发并完善了人工标定数据导出与参数提取系统软件。针对交叉口路段早晚高峰时段不同场景下的交通视频,利用相应算法获取排队长度与停车延误,并将计算结果与人工记录数据进行对比分析,验证基于交通波检测算法获取交通参数的有效性。
[Abstract]:With the continuous and rapid growth of urban car ownership, traffic congestion in urban roads is becoming more and more serious, and a more reasonable traffic control strategy is urgently required. The accurate acquisition of traffic parameters is the premise of optimizing the traffic signal control strategy. Therefore, the use of advanced information technology to accurately obtain traffic parameters can improve urban traffic. In recent years, the intelligent traffic parameter detection algorithm based on machine vision has come into being. Using computer and image processing technology to detect all kinds of traffic parameters and analyze the characteristics of intersection traffic flow has become a hot spot. Especially in the early and late peak period of the intersection, the vehicle queuing is serious and the high precision is realized by video method. The parameters extraction is challenging. In this paper, the traffic parameters acquisition based on traffic wave detection at the intersection is studied in this paper. The main contents include the following aspects: 1. a method of obtaining real data of traffic parameters based on artificial calibration is proposed. The method uses VIPER intelligent software to obtain vehicles at every moment. Pixel coordinates in the frequency image. Using the camera model, the pixel coordinates are converted to the road plane space coordinates. The traffic parameters such as the queue length and the parking delay are determined according to the time and position of the vehicle, and the true traffic wave to the distance curve of the parking line is fitted out according to the starting and stopping changes of all the calibrated vehicles, and the accurate intersection is obtained. The current video based traffic wave detection method is analyzed by.2., which is based on a single camera based complex expansion window algorithm, based on the data fusion algorithm of idols and the process of pixel layer data fusion algorithm, and the traffic wave location data obtained by artificial calibration based on the manual calibration. Three typical traffic wave detection algorithms are compared. Through the establishment of evaluation indexes, the performance of each algorithm is analyzed and the traffic parameters are obtained based on traffic wave detection, including queuing length, parking delay, wave speed and other traffic parameters. The parameters are established by the number of parameters and parking waves and starting waves. The calculation method of the corresponding parameters is derived. The calculation of average parking delay should be based on the existence of only one queue and two queues within the cycle. Based on the experiment, the software of the artificial demarcated data derivation and parameter extraction system should be developed and perfected. In the different scenes of the early and late peak periods of the intersection section, the.4. is developed and perfected. Traffic video, using the corresponding algorithm to obtain the queue length and parking delay, and compare the calculated results with the manual data, verify the effectiveness of the traffic parameters based on the traffic wave detection algorithm.
【学位授予单位】：北京工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491

【参考文献】