| 题名 | 摄像机标定算法研究 |
| 作者 | 李欣菊 |
| 学位类别 | 工学硕士 |
| 答辩日期 | 2003-06-01 |
| 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 | 吴福朝 |
| 关键词 | 摄像机标定 圆环点 准仿射不变量 无穷远平面单应矩阵 运动参数 Camera calibration circular point affine invariance infinite homography motion parameter |
| 其他题名 | Study on Algorithms for Camera Callbration |
| 学位专业 | 模式识别与智能系统 |
| 中文摘要 | 摄像机标定和三维重建是计算机视觉技术研究的重要方向之一，有着很高 的理论研究价值和广泛的应用前景。国内外很多研究者在此领域取得了丰硕的 成果。而本文是在前人的基础上，对摄像机内参数的标定和运动参数的恢复做 了进一步的探索研究，主要工作可以归纳为下： （1） 基于圆的摄像机自标定方法。首先定义平面上两圆的关联直线，再证明的 同一平面上相离的两圆及其关联直线的分布关系为准射影不变量，然后利 用这一性质进行摄像机标定。此方法的特点是基于二次曲线的拟合，不需 要知道图像点和三维空间点的对应关系，也不需要知道图像之间的点对应 的关系。本方法简单易行，且实验证明其鲁棒性较好。 （2） 基于平面相似图像的摄像机标定方法。提出一种基于平面相似图形的摄像 机自标定方法。通过平面相似图形的图像点之间的射影关系，得到该平面 的圆环点的图像，然后根据圆环点的图像对内参数提供的约束，进行摄像 机内参数标定。实验证明本文方法是可行的，并有较好的鲁棒性。 （3） 标定两视点摄像机尺度因子的线性算法。无穷远平面的单应矩阵为两视点 摄像机内参数提供5个非线性约束，我们在假定摄像机畸变因子为零的情 况下，给出了利用摄像机主点计算4个尺度因子的解析表达式。若主点已 知，利用该公式可以直接计算尺度因子，而通常的方法需要求解一个四元 二次方程组。 （4） 恢复摄像机的运动参数的解析公式。提出了从本质矩阵计算运动参数的解 析公式，通过该公式计算的运动参数与传统的SVD分解方法有相同的精 度。 |
| 英文摘要 | Camera calibration is a key step for 3D reconstruction from 2D images. The thesis is concentrated on the exploration of new calibration theories and algorithms. The mains points can be summarized as follows: 1) We develop a new method of camera calibration based on coplanar circles. Firstly, we define the lines associated with two coplanar circles and prove that the distribution of two coplanar circles without real intersections with respect to their associated lines is a quasi-affine invariant. Then the results are applied for camera calibration. This quasi-affine invariant based calibration method has several advantages: (1) it is based on conic fitting, hence inherently more robust; (2) it does not need any image point matching, a widely recognized difficult problem. Experiments with both synthetic data and real images show that it is highly robust. 2) We propose a new technique for camera calibration based on two coplanar similar geometrical entities. Under such conditions, the corresponding image points of the entities are related by a 2D projective transformation, by which the images of the circular points of the plane can be computed, then camera's intrinsic parameters. Experiments show that this calibration method is workable and of good robustness. 3) In the literature, given the infinite homography between two images, the 4 effective focal lengths (fu,fv) and (f'u,f'v) associated with the two cameras both with zero-skew are determined by solving quadratic equations. In this work, we find out analytical expressions of these 4 parameters as functions of the two principal points. One of the advantages of these analytical expressions is that once the principal points are known, they can be directly computed. 4) We propose a new method for analytical solutions of camera motion from the essential matrix. It is shown by extensive simulations and experiments with real images that the proposed method can achieve comparable results in terms of accuracy and robustness with the SVD based method, a widely used one in the literature. |
| 语种 | 中文 |
| 其他标识符 | 688 |
| 内容类型 | 学位论文 |
| 源URL | [http://ir.ia.ac.cn/handle/173211/6846]  |
| 专题 | 毕业生_硕士学位论文 |
| 推荐引用方式 GB/T 7714 | 李欣菊. 摄像机标定算法研究[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2003. |
| 个性服务 |
| 查看访问统计 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
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。
修改评论