【作者】 唐双发；

【导师】 周纯杰；

【作者基本信息】 华中科技大学 ， 控制理论与控制工程， 2009， 硕士

【摘要】 车辆检测是智能交通监控系统的关键环节。随着科技发展和技术进步,车辆检测方式正历经着由传统的线圈检测、雷达检测发展到更为先进的实时视频检测方式。车辆视频检测的基本原理是通过对从摄像机采集的视频序列通过数字图像处理技术,提取图像中的车辆信息,从而达到检测和判断车辆是否违章的目的。视频检测是一个复杂的过程,不仅需要实时采集图像,而且很多图像处理算法不具有通用性,使得系统的开发周期延长,可扩展性不强。鉴于此,本文采用OpenCV作为开发平台,该库在Windows以及Linux下均可以使用,它是开放源代码并且完全免费,这就可以保证库的维护和更新更加便捷和迅速。OpenCV提供了很多基本图像处理方法的函数,并且征对视频图像处理特别提供了几种运动物体跟踪的原型,非常适合于车辆视频检测系统的开发。论文设计并实现了一套基于OpenCV的视频车辆检测算法,算法综合考虑系统检测的可靠性和实时性。在算法流程中的背景初始化与更新步骤中,测试了现有的最常用的三种背景更新算法效果并总结其优劣,同时在此基础上提出了一种背景初始化的方法。在算法流程的另一个关键步骤阴影消除,提出了基于三种色彩空间RGB、YUV、HSV空间的阴影消除算法并进行了实验测试,阴影消除效果明显。在提取车辆特征上,采用状态机的方式对虚拟线圈区域判断当前是否有车辆经过,增强了检测系统的鲁棒性。论文对所进行的一些工作进行了总结,并对车辆视频检测的发展进行了相关预测。更多还原

【Abstract】 Vehicles detection is the key of an intelligent traffic monitoring system. With the development of technology, the methods of vehicles detection are developing from the traditional loop detection, radar detection to advanced real-time video detection. Its basic principle is that via digital image processing techniques extracting the vehicle information from the sequences of video image collected by cameras in order to achieve the purpose of detecting and determining whether the vehicle is illegal.Not only Video detection requires collecting images in real time, but also many image processing algorithms are not universal, which makes the system development cycle to be delayed and the scalability of system to be not strong. In view of this, OpenCV is utilized as a development platform, which can be used under both Windows and Linux. It is an open source and completely free of charge so that the maintenance can be guaranteed and the update of its library can be more convenient and rapider. it provides lots of basic functions of image processing methods as well as several prototypes of tracking the moving objects aim at the video image processing.An OpenCV-based video vehicle detecting algorithm is designed and achieved. In its background initialization and update processes, the existing three most common algorithms are tested to discuss their advantages and disadvantages, and then a new method of initializing the background is brought forward. In another key step of this algorithm, called shadow elimination, new shadow elimination algorithms based on three color space that are RGB, YUV and HSV space are brought forward. Also, we do experiments to show its much better effect on eliminating the shadow. In the features extraction of vehicles, the state machine is used on the virtual loop region to determine whether there are vehicles passing to enhance the robustness of the detection system.The summary of our work are gave, and the development of video detection of vehicles is relevant forecasted.更多还原