基于智能视频分析的特定车辆识别研究.docx

基于智能视频分析的特定车辆识别研究摘 要本课题的出发点是基于现有的、已经覆盖全市几乎所有交通路口、停车场等各类现有视频监控系统，利用路口等处的大量监控探头所提供的视频，运用视频和图像处理分析技术，实现对某些特定车辆的自动识别方法的研究探讨，再通过监控视频的联网，实现对某些特定车辆的位置和行车轨迹等信息进行记录和查询，为某些特定的业务工作快速确定和定位某些特定车辆提供技术手段。在本文中对系统中涉及到的图像预处理、HOG特征提取、特征库建立、SVM分类器的训练，以及对算法的改进都做了详细的论述，同时针对有代表性的算法做了认真的分析学习，主要从以下几个方面开展论文的书写。第一章，主要介绍了国内外在运用视频监控系统对车辆识别检测的研究现状，认识到该课题的研究价值及实际意义；第二章，对图片特征的提取方法做了比较分析，最后选取HOG特征作为车辆的特征；第三章，就目前通用的分类器训练方法做了简要分析，通过比较各自的优缺点，再结合本课题目标对象的特征，选取SVM作为分类器；第四章，对现有HOG特征在灰度域中提取的方法进行改进，将其在颜色域里提取，在HOG特征中含有颜色信息，同时对算法进一步的改进，结合视频的多角度识别处理方法，得到改善后的算法，并实验验证算法有效；第五章，对本课题取得的成果做一总结，并确定之后的研究方向。关键词：特定车辆，车辆识别，HOG特征提取，颜色域HOG特征提取，SVM，多角度识别SPEACIFIC VEHICLE IDENTIFICATION BASED ON INTELLIGENT VIDEO ANALYSISABSTRACTThe point of this project is based on various existing video surveillance systems, has covered almost all the city's traffic junctions and parking lots and so on. The use of a large number of videos which are provided by video surveillance cameras, and the use of video and image processing and analysis techniques, to achieve certain specific vehicle automatic identification method of study was to investigate. Also through surveillance video networking， to achieve certain specific vehicle information such as the location and driving track record and query for specific operational work quickly identify and locate certain vehicles technical means.In this paper, the system involves image preprocessing, HOG feature extraction, feature database creation, SVM classifier training. Also on the improvement of the algorithm has discussed in detail, and has learned and analyzed representative algorithms carefully, mainly from the following aspects to write papers. The first chapter, the research status of use of video monitoring system on detection of vehicle identification at home and abroad was introduced, and knowing the values and practical meanings of the subject; the second chapter, analyze various picture features extraction methods, finally select HOG features as vehicles features; the third chapter, ana