【作者】 李小峰；

【导师】 李峰；

【作者基本信息】 江苏大学 ， 计算机应用技术， 2008， 硕士

【摘要】 实时交通流数据的采集在智能交通系统中起着重要的作用。交通流检测有多种方式,而基于图像处理的视频检测方式近年发展很快,它具有大区域检测、设置灵活等优越性,已成为智能交通系统的一个研究热点。本文致力于研究一种灵活可靠的视频交通流检测系统,将智能传感器的概念引入视频检测中,以提高检测效率。车辆的检测基于车道,在每个车道设置两组虚拟检测线(每组虚拟检测线包括横向和纵向虚拟检测线)来检测交通流参数,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转化成一维的检测信号,减小了运算量,降低了运算负荷。该方法的特点是处理运算的图像区域小,避开了在二维图像空间中进行车辆跟踪。通过大量的实验和分析,论文完成了复杂道路交通场景下的车辆检测和车速测量算法的开发工作,并利用Matlab进行了仿真和验证。论文的主要工作包括:(1)改进了虚拟检测线的设置,每组虚拟检测线由横向和纵向两部分组成,在横向和纵向虚拟检测线上分别通过帧间差分和背景差分进行特征信号的提取。在分析虚拟检测线位置和大小对输出信号的影响基础上,提出虚拟检测线设置中应遵循的一些原则。(2)采用形态除噪方法、Kirsch边缘检测方法、基于Kalman的背景更新方法以及基于结构单元的阴影去除方法进行图像预处理。(3)在交通流信息采集系统中为了精确地在一维信号中跟踪车辆,系统引入了正负信任帧、正负信任段等重要参数及奖惩策略,同时为了提高速度测量的精度和减小计算量,系统利用车头分别进入上下两横向检测线时候的特征信号求取车辆的速度。(4)利用基于ARM-Linux的嵌入式平台完成交通流检测原型系统的设计,验证算法的实时性和方便性。实验证明本文提出的基于视频的交通流信息采集方法具有一定的稳健性,实时性,方便性和鲁棒性。更多还原

【Abstract】 The collection of real-time traffic data plays a critical role in intelligent transportation system.Technological innovations have given rise to different types of traffic detectors. Meanwhile,a promising approach,video-based measurement system,has developed quickly.Since it has many advantages,for instance,wider-area detection and superior flexibility,many researches have been done in this area.In this paper,in order to develop a flexible and reliable system for detecting the traffic parameters through image sequences,the intelligent image sensors is present to improve detecting efficiency.The vehicle detection is based on the roadway,which sets two virtual loops(every virtual loop contains a transverse area and a lengthways area),like the inductive loop sensor,to detect their traffic flow parameters.The key idea of the system is converting the two-dimensional image data to one-dimensional digital temporal signals by the virtual loop,which reduces computation load.Vehicle detection and volume measurement algorithms have been accomplished through a large number of experiments and plentiful analysis in the traffic scene that can be simple or complicated.And algorithms have been emulated and testified by MATLAB.Contribution in this dissertation as follow:(1) The virtual-loop set in the image has been improved,which contains a transverse area and a lengthways area.Each virtual-loop’s output signals mainly derive from the pixel difference between consecutive image frames within the transverse virtual-loop area and the pixel cost of background subtraction image within the lengthways virtual-loop area.When the virtual-loop was set,some rules about position and size of the virtual-loop have been present.(2) In the previous management of image,the mathematical morphology method,the Kirsch edge detection algorithm,the Kalman algorithm selected to updated the background within the lengthways virtual-loop area and the shadow eliminating method based on structural elements were used.(3) In order to tracking the vehicle more inerrably,system uses some important parameters such as positively and negatively credible frame、positively、negatively credible segment and so on,and an effective algorithm(punishment and reward algorithm) was present.In order to measure the volume of vehicles more inerrably and easily,the output signal produced when the head of the vehicles drove into the two transverse virtual-loop area in one lane was used.(4) A prototype system was also designed on embedded arm-linux and proved the system was real timely and convenient.Experiments have proved that the method of traffic information detection based on video was simple and available,real timely,convenient,with good tolerance and applicability.更多还原