【作者】 唐国侠；

【导师】 李扬；

【作者基本信息】 广东工业大学 ， 信号与信息处理， 2008， 硕士

【摘要】 随着社会经济的快速发展,交通系统日益复杂。为了给交通系统的管理提供各种实时交通信息,以方便、高效地利用和管理现有的交通系统,基于视频图像处理的交通信息采集系统成为研究的一个热点。本课题作为交通信息采集系统的一个重要子系统,它以实时的车辆行驶速度为研究检测对象,其目的是改进实时车辆速度的采集手段,为道路交通管理提供更为有效的依据,提高交通管理的自动化、现代化水平。本文致于研究一种灵活、可靠的视频交通流测检测系统。车辆的检测基于车道,在每个车道可以设置一条虚拟检测线和一个虚拟检测区域来检测交通流的车速,虚拟检测线的作用类似于电磁感应线圈。系统通过对视频虚拟检测线的预处理将二维的数字图像信号转化为一维的检测信号。该方法的特点是只需对虚拟检测线区域内的图像进行处理,处理运算的图像区域小,因此减少了运算量,降低了运算负荷。本文还提出了一种从频率域出发,估计出运动车辆在空间域上位移的算法。利用傅里叶变换的自配准性质和位移特性,用极坐标形式下连续两帧车辆图像相位谱的差可直接估计运动车辆的位移,进而估计出车速u=s/△t。由于本算法分辩能力不低于一个像素,因而可用于车速的高精度测量。利用本文提出的车速测量算法,对晴天和阴天两种典型天气条件下的交通流视频进行了实验,实验结果表明具有较高的车流量和车速检测精度。更多还原

【Abstract】 With the rapid development of society and economy,the traffic system becomes more and more complex.In order to provide all kinds of real-time traffic information to the management of traffic system for achieving efficient management and making full use of the traffic system,the study on real-time traffic information collection system based on video image processing has been becoming a focus. As an important subsystem of real-time traffic information collection system, the paper makes the running speed of vehicles as the studying and detecting object and mainly studies on the detecting algorithms of real-time speed and other related ones. The purpose of this subject is to improve the detecting methods of real-time speed and provide more efficient references to increase the automatic and modernizing traffic management levels.This paper focuses on developing a flexible and reliable system to detect the real-time speed of the traffic flow parameters through image sequences. The vehicle detection is based on the roadway, which sets a virtual line and a virtual-loop area, like the inductive loop sensor, to detect the traffic flow parameters. The key idea of the system is converting the 2-dimensional digital image to 1-dimensional detection signal by virtual detection line preprocessing, which reduces computation load. The trait of this method is only processing little image area within virtual line or virtual-loop and avoiding vehicle tracking in 2-dimensional image, hence the time cost of calculation and the computation burthen is recuced.An algorithm which is used to estimate the translation of a moving object in spatial domain according to its Fourier transform spectrum is also proposed in the paper. Based on shift theorem of Fourier transform and auto-registration, the algorithm directly estimates the translation with the phase spectrum difference between continuous images of a moving object in polar coordinate system. Vehicle velocity v=s/△t (where s is the displacement of vehicle and At is the interval between two continuous images) is calculated corresponding. It can give the displacement of a moving target with a resolution less than 1 pixel, hence it can be used to measure vehicle velocity with high precision.Using the algorithms put forwarded in this paper, the traffic flow video files in fine day and cloudy day are experimented and the experimental results show that the detection accuracy is rather high.更多还原