面向视频交通流参数的摄像机动态标定技术研究
文献类型：学位论文
| 作者 | 李元涛 |
| 学位类别 | 工学博士 |
| 答辩日期 | 2007-06-07 |
| 授予单位 | 中国科学院研究生院 |
| 授予地点 | 中国科学院自动化研究所 |
| 导师 | 王飞跃 |
| 关键词 | 智能交通系统 交通视频系统 摄像机动态标定 车道线检测 交通流参数 intelligent transportation system video-based traffic management system dynamic camera calibration lane detection traffic flow parameters |
| 其他题名 | Video Traffic Flow Parameters-Oriented Dynamic Camera Calibration Technology Research |
| 学位专业 | 控制理论与控制工程 |
| 中文摘要 | 随着经济的高速发展、城市化进程加快和汽车数量急剧增长，导致交通拥挤、事故频发、环境恶化和能源紧张等日益加剧的问题，从而产生了运用信息技术手段来解决上述问题的智能交通系统。图像处理与分析在智能交通领域扮演着极其重要的角色，具有广泛的应用价值，尤其是摄像机已经广泛应用于交通视频系统。基于视频的交通系统基本功能是通过图像处理技术自动地提取实时的交通流参数，而交通流参数获取的准确度完全取决于摄像机参数标定的精确度和图像处理算法运用的是否恰当。因此，只有摄像机参数被正确地标定，基于视频的交通流参数才能准确的获得，所以本文主要研究面向视频交通流参数的摄像机动态标定技术。动态标定技术的优点在于能够及时标定摄像机以满足交通流参数检测的准确性要求，本文围绕着这个研究方向主要分为以下几个方面：1. 分析了国内外摄像机标定的研究现状，以及智能交通视频系统的研究发展，指出当前该领域中摄像机标定研究存在的问题和解决方案。2. 建立了一种基于交通视频场景的摄像机标定数学模型。主要利用交通场景中普遍存在的车道线平行特征、和投影几何学中平行线性质来建立数学模型，并且推出针对该模型给出了易于求解摄像机参数的方法。3.综合现有算法给出了一套适合摄像机标定的车道线检测算法。主要围绕本文建立的摄像机标定数学模型，综合考虑影响车道线检测准确性的各个因素，集成现有算法给出一套适合摄像机标定的自动的图像分析与处理算法。4.提出了基于长度和位移的视频场景中的交通参数获取方法。主要围绕摄像机标定来准确地求取交通流参数，重点放在与长度或位移相关的数据计算方面。5.设计了一种面向交通视频场景的摄像机动态标定系统，并对技术难点进行了分析说明。主要从车道线检测算法和摄像机参数校正两个模块，系统的自动化和动态性两个方面，分析和设计该系统的框架和结构。6.通过实验验证了本文提出的摄像机标定数学模型、车道线检测算法以及交通参数获取方法的有效性和技术实现的可行性。 |
| 英文摘要 | With the rapid development of economy, the acceleration of urbanization and the generalization of automobiles, many problems have been becoming more serious, such as traffic congestion, frequent accidents, environment deterioration and energy shortage. So intelligent transportation system (ITS) have come into being though utilizing information technique in order to solute the above problems. Image processing and analysis have played an important role inintelligent transportation system, possessing extensive application value. Especially camera have been widely applied to traffic management system(TMS). For a video-based TMS(vTMS), the basic function is automatically to extract real-time traffic parameters, the accuracy of which is depended on the calibrated camera parameters and the used image processing algorithms. So if only camera parameters can be calibrated properly, video-based trafficparameters can be estimated accurately. Therefore, this thesis has mainly researched dynamic camera calibration technology oriented on video traffic flow parameters, which can satisfy the requirement of the exactness of the detectedtraffic parameters by adjusting camera parameters. With this research object, the contributions of the thesis include the following aspects:1.This thesis analyzes research current status and development of camera calibration, and points out existent problems and solutions of camera calibration in the fields of current vTMS.2.This thesis establishes a mathematical model of camera calibration based on traffic video scenes. It mainly makes use of the characteristic of ubiquitous parallel lane markings and the character of parallel lines in projective geometry, in addition deduces the solution approach of camera parameters according to this model.3.This thesis presents a set of lane detection algorithms by integrating a few of existing algorithms in traffic video scenes, subsequently synthesizes and provides a series of image processing and analysis algorithms applied to dynamic camera calibration, synthetically considering various factors which affect the exactness of lane etection.4.This thesis researches a few methods of acquiring traffic parameters about length or location in video scenes. It mainly encircles the calibrated camera parameters to obtain traffic parameters, and the interrelated emphasis is length measurement.5.This thesis presents a dynamic camera calibration system oriented on traffic video parameters. It mainly analyzes and designs the synthetical solution of which from four aspects, such as lane detection algorithm, camera parameters calibration, systemic automation and dynamic.6.This thesis makes some experiments in four aspects, such as mathematical model of camera calibration, lane detection algorithm and acquiring traffic parameter method. |
| 语种 | 中文 |
| 其他标识符 | 200418014628064 |
| 源URL | [http://ir.ia.ac.cn/handle/173211/6008]  |
| 专题 | 毕业生_博士学位论文 |
| 推荐引用方式 GB/T 7714 | 李元涛. 面向视频交通流参数的摄像机动态标定技术研究[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2007. |
入库方式： OAI收割
来源：自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。