摘要
针对单幅图像提取三维信息困难,设计了一种基于单幅图像的三维信息提取算法。该算法首先需要标定摄像机,得到摄像机从三维到二维的成像矩阵M3×4,并计算M3×4的伪逆矩阵M+4×3,其本质是通过最小二乘法求解矛盾方程组;利用M+4×3提取二维单幅图像中的三维信息,再把这些三维信息投影到一个平面上显示出来。实验表明,单幅图像目标物体必须满足共面等约束条件,从而可以对目标物体的三维信息做定性分析(定性视觉)。此外,还对三维空间投影到二维图像的过程进行了理论推导,得出了如果没有足够的先验知识,成像过程不可逆的结论。
A 3D information extraction algorithm based on a single image was designed in this paper for the difficulties of this research field, in this algorithm, camera imaging matrix M3x4 was obtained by camera calibration firstly, the pseu do inverse matrix M4X3 was calculated by M3x4,the essence of this computation was solving inconsistent ectuations by least squares approximation, 3D information would be extracted from a single image through M1x3,the result wasshown in a image by projection operator. The result could be used for qualitative analysis (qualitative vision) currently due to the limited information from a single image. The experiment indicates that the extracted object should be coplanar. Moreover, the essential theories of 3D to 2D's projection were deduced, the conclusion of the deduce is that imaging process is an irreversible process.
出处
《计算机科学》
CSCD
北大核心
2010年第5期243-246,共4页
Computer Science
基金
国家自然科学基金(60705005)
博士点新教师基金(20070610031)
四川省应用基础研究计划(2008JY0038)
教育部新世纪人才支持计划(NCET-08-0370)资助
关键词
伪逆矩阵
三维信息提取
单幅图片
定性视觉
Pseudo inverse matrix 3D information extraction Single image Qualitative vision