CASIA OpenIR  > 毕业生  > 硕士学位论文
基于视频分析的车辆违章检测技术研究
Alternative TitleResearch on Video Analysis based Vehicle Violation Detection Technology
刘振华
Subtype工程硕士
Thesis Advisor刘昌平
2012-05-25
Degree Grantor中国科学院研究生院
Place of Conferral中国科学院自动化研究所
Degree Discipline计算机技术
Keyword车辆检测 车辆跟踪 车辆违章检测 支持向量机 Vehicle Detection Vehicle Tracking Vehicle Violation Detection Support Vector Machine
Abstract如何对车辆违章进行有效检测这一问题一直是交通管理部门关注的重点。传统的车辆违章检测方法基本上都是基于电磁感应原理来工作的,该方法的缺点是可靠性低且安装与维护不方便。随着计算机视觉和数字图像处理技术的发展,以及智能交通系统(Intelligent Transportation System, ITS)理念的提出,研究人员提出了基于视频分析的车辆违章检测方法,该方法大致包括三个主要的模块:车辆检测、车辆跟踪和行为理解,能够自动对车辆违章行为进行检测,与传统方法相比,该方法具有很多优势。基于此,本文对目标检测与跟踪算法进行了研究,提出了新的车辆检测算法与车辆跟踪算法,并在此基础上实现了车辆违章检测算法。本文的主要工作和创新点包括: (1) 研究了目标检测算法,分析了现有的基于视频序列的目标检测算法和基于静止图像的目标检测算法。针对车辆这一特定目标,提出一种由粗到细的双层车辆检测算法,首先利用训练得到的车辆的边缘模板进行全局搜索,给出车辆的大致位置;之后用事先训练好的车辆分类器模型对粗定位结果进行二次检测,实现对车辆的准确稳定检测。 (2) 对现有的目标跟踪算法进行研究,提出了一种基于支持向量机(Support Vector Machine, SVM)的车辆跟踪算法,首先根据车辆在上一帧中的位置确定出车辆在本帧中的候选区域;然后用车辆检测环节中训练出来的SVM分类器模型对候选区域内的每一个位置进行验证,得出车辆的最佳位置;同时对交通视频场景进行背景建模,利用运动前景检测结果对车辆的最佳位置进行修正,得到最终跟踪结果,从而实现对车辆的快速有效跟踪。 (3) 实现了违章变道行驶检测算法、违章逆行检测算法和违章停车检测算法,算法均包括车辆检测、车辆跟踪和违章行为判别三个基本模块。违章变道行驶检测算法根据车辆的行驶状态与初始所属车道是否一致来判断是否有违章行为发生,违章逆行检测算法和违章停车检测算法是建立在对车辆运动轨迹进行分析的基础上的,可以对违章变道行驶、违章逆行和违章停车这三种违章行为进行有效检测。
Other AbstractHow to detect vehicle violation effectively has been the focus of the traffic management department. Conventional vehicle violation detection methods are basically based on the principle of electromagnetic induction to work and have the shortcomings of low reliability, inconvenient installation and maintenance. With the development of computer vision and digital image processing technology and the appearance of the concept of Intelligent Transportation System, researchers have proposed a vehicle violation detection method based on video analysis which generally consists of three main modules, that is vehicle detection, vehicle tracking and behavior understanding, vehicle violation can be detected automatically and compared with conventional methods, this method has many merits. So this paper makes some research on object detection and tracking algorithm and new vehicle detection algorithm and vehicle tracking algorithm have been proposed, and based on the vehicle detection and tracking algorithm, a vehicle violation detection algorithm has been realized. The main work and innovation are as followed. (1) Introduce the existing object detection algorithm based on static images and video sequences. A two-level vehicle detection algorithm has been proposed. Firstly, a globally search has been made which uses the edge template trained before to give the approximate location of the vehicle. Secondly, with the pre-trained vehicle classifier model, a second detection on the coarse locations given by the first step has been made to achieve an accurate and stable vehicle detection results. (2) Introduce the existing object tracking algorithm. A vehicle tracking algorithm based on support vector machine has been proposed. Firstly, vehicle's candidate region in this frame is identified based on vehicle's position in the last frame. Secondly, each position within the candidate region is validated using the support vector machine classifier model trained in vehicle detection part to get the best position of the vehicle. Background model of the traffic video scene is achieved simultaneously and vehicle’s best position is corrected by using the result of motion foreground detection to get the final tracking result which is robust and can be achieved quickly. (3) Vehicle violation detection. Three basic modules of vehicle violation detection algorithm are vehicle detection, vehicle tracking and violation behavior identification. A judgment that whether lane violation...
shelfnumXWLW1792
Other Identifier2009M8014628004
Language中文
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/7637
Collection毕业生_硕士学位论文
Recommended Citation
GB/T 7714
刘振华. 基于视频分析的车辆违章检测技术研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2012.
Files in This Item:
File Name/Size DocType Version Access License
CASIA_2009M801462800(3571KB) 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[刘振华]'s Articles
Baidu academic
Similar articles in Baidu academic
[刘振华]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[刘振华]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.