基于高速公路大货车违法占道监测系统的车辆检测与跟踪研究

发布时间：2018-02-15 01:54

本文关键词： 视频车辆检测光照阴影消除运动位置跟踪车型识别　出处：《西南交通大学》2015年硕士论文　论文类型：学位论文

【摘要】：随着社会经济的飞速发展以及交通运输量的日益提高,传统的交通监管方式已不再能满足复杂多变的交通现状,智能交通系统应运而生。本文课题来源于四川省科技厅科技支撑项目“智能交通安全监测与流量控制系统重大关键技术研究”项目的子课题一：大型货车占道行驶动态检测关键技术研究及监控系统开发。针对高速公路这一特殊场景,研究车辆检测与跟踪、车型识别的方法,使能够有效检测和跟踪高速公路行驶车辆,准确识别大型货车,为建立大型货车违法占道判别模型提供有力支撑。首先详细讨论了各种道路信息采集方式的原理、应用场景,并对各种道路信息采集方式的优缺点进行了比较,选择了监控效果直观清晰、安装维护方便、监控范围广的视频信息采集方式。分析了目标检测、跟踪以及车型识别的基本方法,对各类方法进行了原理介绍和优劣比较。其次提出了在时空背景差分模型的检测基础上,加入了光照阴影消除和运动种子填充和运动目标尾迹消除的新检测方法,该方法使检测目标更加清晰完整。最后将该检测方法进行了相应实验,并将该方法的车辆检测效果和混合高斯检测模型进行了二值差分图的对比,实验结果表明该方法能达到较好的目标正确识别率,同时,还将错误样本进行了归类和错误原因分析。结果证明,在不同密集程度的视频中,车辆检测的正确率都能达到90%以上。另外,根据高速公路上车辆运动轨迹方向基本不变的特性,采用了基于车辆的运动位置信息和颜色信息关联相结合的车辆跟踪方法,详细介绍了跟踪原理。最后通过分析车辆面积在视频中的非线性关系设定阂值以及车辆本身的纹理信息和轮廓信息,实现了大小车区分和客货车区分。通过选择大量样本对本文中的车辆检测跟踪及车型识别方法进行实验。证明可以应用在大型货车占道行驶动态检测关键技术研究及监控系统开发中,为建立违法占道判别模型提供依据。
[Abstract]:With the rapid development of social economy and the increasing traffic volume, the traditional traffic supervision can no longer meet the complex and changeable traffic situation. Intelligent transportation system arises at the historic moment. This paper comes from the project of science and technology support project of Sichuan science and technology department, "research on the key technology of intelligent traffic safety monitoring and flow control system" Research on key Technologies of driving dynamic Detection and Development of Monitoring system. The method of vehicle detection and tracking and vehicle type identification is studied so as to effectively detect and track motorway moving vehicles and accurately identify large trucks. This paper provides a powerful support for the establishment of the model for judging the illegal occupation of large freight cars. Firstly, the principle and application scene of various road information collection methods are discussed in detail, and the advantages and disadvantages of various road information collection methods are compared. The method of video information acquisition with visual effect, convenient installation and maintenance and wide monitoring range is selected. The basic methods of target detection, tracking and vehicle identification are analyzed. The principle of each method is introduced and the advantages and disadvantages are compared. Secondly, a new detection method of light shadow elimination, moving seed filling and moving object wake cancellation is proposed based on the space-time background difference model. This method makes the detection target more clear and complete. Finally, the corresponding experiments are carried out, and the vehicle detection effect of this method is compared with the mixed Gao Si detection model. The experimental results show that the method can achieve a good target recognition rate. At the same time, the error samples are classified and the error causes are analyzed. The accuracy rate of vehicle detection can reach more than 90%. In addition, according to the characteristics of the moving track direction of the freeway, the vehicle tracking method based on the association of the vehicle motion position information and the color information is adopted. The tracking principle is introduced in detail. Finally, by analyzing the nonlinear relationship between the vehicle area and the video, the threshold value, the texture information and the contour information of the vehicle itself are analyzed. By selecting a large number of samples, the vehicle detection and tracking and vehicle identification methods in this paper are tested. It is proved that it can be applied to the research of key technologies of dynamic detection of large freight cars on the road. Research and monitoring system development, It provides the basis for establishing the discrimination model of illegal occupation.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;TP391.41

【参考文献】