节点文献
运动参数估计与图像校正
ESTIMATION OF MOTION PARAMETERS AND IMAGE RECTIFICATION
【摘要】 本文通过理论分析,严格证明了本质矩阵通过奇异值分解可以唯一地获得旋转参数两个可能解,同时利用单位球进一步给出并证明了两个可能解直观的几何描述。同时将这种运动参数估计方法引入立体视觉中图像校正等问题,所给出的校正算法能够保证校正后的图像满足透视几何约束。
【Abstract】 The uniqueness of the motion parameters is discussed though the Essential matrix SVD decomposition. It has been proved that the two candidates of rotational matrix could be uniquely determined by computing the SVD decomposition of Essential matrix. And the intuitionistic geometric descriptions of the two candidates are given though the unit sphere. The given method for estimating the motion parameters is introduced to the image rectification for stereo vision. The presented algorithm can ensure that the rectified images preserve perspective geometry.
【关键词】 本质矩阵;
奇异值分解;
单位球;
图像校正;
【Key words】 Essential Matrix; SVD Decomposition; Unit Sphere; Image Rectification;
【Key words】 Essential Matrix; SVD Decomposition; Unit Sphere; Image Rectification;
【基金】 国家自然科学基金(No.60203003);清华大学基础研究基金资助项目
- 【文献出处】 模式识别与人工智能 ,Pattern Recognition and Artificial Intelligence , 编辑部邮箱 ,2003年01期
- 【分类号】TP391.41
- 【被引频次】9
- 【下载频次】143