节点文献

基于单目视频的车辆对象提取及速度测定方法研究

【作者】 张帆

【导师】 李勃;

【作者基本信息】 昆明理工大学 , 计算机应用技术, 2014, 硕士

【摘要】 随着国民经济水平的稳步提高,国家高速公路网建设进程也在不断加快,使人们的出行越来越便利。但与此同时,为了保证高速公路的正常秩序,保护人民生命财产安全,公安交警部门也迎来了巨大的挑战。智能交通系统(ITS)的出现,大大减轻了交通监管部门的人力物力,降低了执法难度,提高了违章肇事的处理效率。机器视觉技术作为智能交通系统的支撑,成为当今的研究热点之一。目前高速公路视频卡口配有违章拍照系统,可以自动识别车辆的牌照,从而在车辆管理处找到登记的车主信息,做到执法有据。但一些不法分子采用遮挡、污损牌照的方式或者使用套牌,使现有的违章拍照系统无法准确识别出车辆信息,只能通过人工的方式进行筛选,需要花费巨大的人力资源。本文的研究内容是分布式视频车辆统一性时空关联检索的前期工作,为下一步的车辆识别、同一性时空关联分析提供支持。本文的研究内容有以下几个方面:1、道路模型的建立。根据高速公路卡口、道路的不同情况,建立起对应的道路模型。道路模型数据在车辆同一性时空关联检索的各个阶段都是非常重要的信息。2、视频中车辆对象的提取。基于混合高斯背景建模方法,对传统背景差分法进行改进,使用基于区域比较的方法,从交通视频中提取车辆对象。使用基于HSV颜色空间的阴影检测与消除算法去除车辆的阴影,得到清晰的车辆对象,满足后续研究要求。3、运动车辆的跟踪。改进了基于图像金字塔Lucas-kanade光流法,使用基于区域特征点的方法对车辆进行跟踪,获得运动轨迹,为车辆速度测定提供依据。4、基于视频的运动车辆速度测定。采用基于高速公路车道线的方法测量车辆的行驶速度。实验结果表明该方法简便易行,准确率高。

【Abstract】 With the increment of people’s incoming, the process of national high way network construction is also rapidly develop. People feel more and more convenience in traffic. However, in order to keep the high way order, protect people’s life and property security, high way traffic administration face a big challenge. Since the Inteligence Traffic System appears, greatly reducing the traffic control department of manpower, reduce the difficulty of enforcement, improve processing efficiency violation accident. The machine vision technology as the support of intelligent transportation system, become one of the hotspot of research.The content of this paper is the prophase work of distributed video vehicle unified spatiotemporal Association retrieval, provide materials for vehicle identification, the back of the same temporal and spatial correlation analysis. The research contents of this paper are as follows:1The establishment of road model. According to the different situation of highway bayonet, and road conditions, establish the corresponding road model. Model data is very important information in each stage of the vehicle identity spatiotemporal Association retrieval.2Extract the vehicle object from the video. Using the modeling method based on mixed Gaussian background model from the highway video to extract vehicle objects, including the shadow elimination. The purpose is to make the vehicle object extraction to meet the following requirements of feature extraction.3Tracking of moving vehicles. Using the improved optical flow based on image Pyramid Lucas-kanade to track vehicle, obtain its trajectory, preparing for the vehicle speed determination.4Vehicle speed measurement based on video. Using the speed measurement method of vehicle based on highway lane. The experimental results show that the method is simple, accurate.

节点文献中: 

本文链接的文献网络图示:

本文的引文网络