【作者】 刘肃亮；

【导师】 周明全；

【作者基本信息】 西北大学 ， 计算机软件与理论， 2004， 硕士

【摘要】 本文设计开发了一个新型的交通车辆违章智能视频监控系统(ITPVMS)，它采用基于交通动态视频的软件检测模式，实现了对交通车辆违章行为的多项目、大范围监控，具有实时性、准确性、高智能性和多功能性等优点，是对现有智能交通监控技术的重大改进。本文的研究工作是公安部部级重点课题“区域交通控制可视化数据处理平台的研究”的重要组成部分。 首先，设计了适当的方案，通过携带高分辨率CCD摄像头的摄像机和视频捕获卡实现对动态交通视频流的实时自动采集； 第二，兼顾系统的实时性和准确性原则，设计快速高效的算法，对视频图像进行预处理。首先运用中值滤波法对视频图像进行去噪处理，然后采用二阶差分法进行图像锐化，再通过视频捕获卡携带的二次软件开发包SDK-2000进行对比度增强，从而得到比较理想的图像，为后续的图像分析工作做准备； 第三，根据系统特点，通过背景图像差分法，分割出监测区中的运动目标，然后对交通目标的区域面积、长度和宽度等进行特征提取，再经过尺度滤波处理，从而自动识别出机动车辆目标： 第四，分别设计相应的算法，实现对闯红灯、压黄线、违章停车、违章超速及违章逆行等常见的交通车辆违章行为的自动识别，并自动抓拍违章现场，截取违章视频，采集其它相关违章信息，为交通执法部门进行交通违章处罚提供有力依据： 考虑到特殊路段特殊环境下的交通复杂性，为了有效监控其它类型的交通违章行为，本文还设计并实现了对交通违章的手动监控模式。更多还原

【Abstract】 In this paper, we introduce the design and development of a new intelligent traffic peccancy monitoring system by video detection (ITPVMS), which adopts software mode based on detecting dynamic traffic video stream. This system could carry out efficient real-time monitoring on quiet a few sorts of traffic peccancies, with merits of high precision, more functions and high automatization . Comparing with the current traffic monitoring techniques, it’s a great improvement. Our research is also one of the most important parts of the project "Research on Visual Data-processing Platform of Regional Traffic Controlling", which is funded by Police Ministry of China.Firstly, a proper scheme is successfully advanced to collect real-time digital traffic video stream automatically, which is based on the cooperation of video-collecting card and vidicon equipped with color CCD camera with high differentiation ratio.Secondly, with a view of real-time and precision of the system, some rapid and efficient algorithms are designed to make image preprocessing on video images. The first step is to eliminate all kinds of noises generated in the process of collecting traffic video by means of median filter; Then we finish image sharpening with Laplace operator to clarify edges of the video images; finally, image contrast is enhanced with the help of the second-developing packet SDK-2000, which is provided by our video-collecting card. This way, much more satisfactory images are obtained, and preparation is already made for the following work, that’s image analyse.Thirdly, according to the characteristics of our system, mobile objects are distinguished by means of image difference. Then some features such as regional area, length and width of traffic objects are selected and extracted. After processing of scale filter, we successfully recognize those moving vehicles appearing in the monitoring regions.Fourthly, some corresponding algorithms are brought forward to automatically identify those familiar traffic peccancies of many sorts, such as red light violation, yellow-line covering, parking in prohibited spots, over-speed running and converse running. Then we find means to snap pictures and to intercept videos of the spot, and collect a lot of relevant information such as the time, the block, the running direction and the weather conditions and so on. According to this information, our traffic management departments could make correct decisions on punishing those drivers who have broken the traffic rules.Considering the complexity of traffic environments on particular blocks, in order to monitor other traffic peccancies effectively, we also have designed the scheme to carry out traffic monitoring by manual.更多还原