CORC  > 自动化研究所  > 中国科学院自动化研究所  > 毕业生  > 硕士学位论文
题名摄像机定标方法研究
作者杨长江
学位类别工学硕士
答辩日期1999-06-01
授予单位中国科学院自动化研究所
授予地点中国科学院自动化研究所
导师胡占义
学位专业模式识别与智能系统
中文摘要摄像机定标是计算机视觉中的一项最基本的任务,本文对这一问题进行了较为深 入的研究,主要工作可以归纳为以下四个方面: 1.基于主动视觉的摄像机自定标方法研究。该方法和文献[1]提出的方法相比较 最大的优点在于不要求摄像机在三维空间作任意方向的平移运动。在适当调整摄像 机俯仰、扫视角的情况下,只需要控制摄像机在同一平面作四组平移运动,其中每 组包括两次相互正交的平移运动,可以线性求解摄像机内参数。理论上严格证明了 如果摄像机作三次俯仰,或者作两次俯仰和一次扫视来调整摄像机平移运动的姿 态,则解存在而且唯一。通过稳定性分析证明前一种方法对误差很敏感,后者则是 稳定的。 2.基于二次曲线的纯旋转摄像机自定标方法研究。摄像机只绕光心旋转,没有平移运动,在不同的方位拍摄两个二次曲线、或两个二次曲面、或一个二次曲线与一 个二次曲面的三幅以上图像,利用图像之间的二次曲线对应,可以唯一地确定摄像 机内参数矩阵,同时可以获得摄像机不同方位之间的旋转矩阵。整个求解过程中只 涉及线性代数中的矩阵处理技术,解是闭合的,不需要迭代过程。 3.基于二次曲线的桌面视觉系统定标方法研究。在Zhang[2]的桌面视觉系统摄像 机定标方法的基础上,我们提出了一种基于二次曲线对应的桌面视觉系统摄像机定 标方法。该方法只需要摄像机在2个或2个以上不同的方位拍摄一个平面模板的图 像,平面模板上包括3个或3个以上的同心二次曲线,摄像机和平面模板都可以自由 移动。由于二次曲线是一种全局化基元,可以极大地降低确定基元对应关系的难 度。所提出的方法在求解过程中不需要非线性迭代,可以直接获得闭合解。该方法 对设备的要求比较低,方法简单,精度和稳定性高,非常适合对计算机视觉不太了 解的普通人员应用。 4.利用球面定标摄像机方法研究。我们提出了通过一个球面定标摄像机的方法。 该方法对设备的要求非常低,只需要一个球的一幅图像就可以确定摄像机的主点位 置,同时在相差一个尺度因子的意义下,可以确定摄像机的横向与纵向等效焦距以 及倾斜因子。如果在不同的方位再取一幅图像,并且求出该图像与前一幅图像之间 的基本矩阵,或者如第五章所述,利用一个平面模板,求出该模板和图像之间的 Homography,那么可以进一步完全确定摄像机的参数。整个求解过程中不需要迭代 过程,通过矩阵计算可以直接获得闭合解。
英文摘要Camera calibration is an indispensable step in 3D computer vision in order to extract metric information from 2D images. In this work, the following four topics are investigated: 1. Camera Self-Calibration for Active Vision Systems In this part, a new camera intrinsic parameters self-calibration technique for ordinary active vision system is proposed. Unlike the one presented in [1], this approach does not require camera of the ability to translate along arbitrary three dimensional directions. By controlling a pan-tilt-translation camera platform to translate along two orthogonal directions four times, while adjusting camera poses, the camera intrinsic parameters can be determined linearly. It is proved that the solution is existing and unique if the camera pose is adjusted by either three tilts or two tilts and one pan. In addition, based on extensive simulations of stability analysis, it is shown that the first one is numerically unstable and sensitive to noise, whereas the second one is robust. 2. Self-Calibration of Rotating Cameras Using Conic Correspondences In this part, instead of corresponding points which are widely used in the literature, we use corresponding conics to calibrate a rotating camera. The corresponding conics can be images of either a planar conic or silhouettes of a conicoid. We show that at least three images, taken from the same position with different camera orientations, of two conics, or two conicoids, or one conic and one conicoid, are needed, and a closed-form solution can be obtained. Our new approach requires no knowledge of the orientations of the camera, and is based only on the image correspondences. 3. Planar Conics Based Camera Calibration Inspired by the technique proposed by Zhang in [2], we proposed a new easy camera calibration technique in this part. It is well suited for users without specialized knowledge of 3D geometry or computer vision. The technique only requires the camera to observe a planar pattern shown at a few (at least two) different orientations. The model plane contains a few (at least three) con-centric conics. Either the camera or the planar pattern can be freely moved. The motions need not be known. All computations involved are matrix operations in the linear algebra, and a closed-form solution can be obtained. Both computer simulation and real data are used to test the proposed technique, and very good results are obtained. Compared with the classical techniques where an expensive calibration pattern is always used, our technique is easy to use and more flexible. 4. Camera Calibration with A Sphere In this part, we present a flexible new technique for camera calibration which uses a sphere as the calibration pattern. In this technique, a single image of the sphere is sufficient to uniquely determine the position of the camera principal point, and to determine the other three camera intrinsic parameters up to a common scale factor. In addition, if tw
语种中文
其他标识符531
内容类型学位论文
源URL[http://ir.ia.ac.cn/handle/173211/7280]  
专题毕业生_硕士学位论文
推荐引用方式
GB/T 7714
杨长江. 摄像机定标方法研究[D]. 中国科学院自动化研究所. 中国科学院自动化研究所. 1999.
个性服务
查看访问统计
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

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。


©版权所有 ©2017 CSpace - Powered by CSpace