摘要
本文介绍了基于奇异值分解的射影重构算法的一般框架,以测量矩阵的秩为4作为约束,以仿射投影逼近透视投影,利用共轭梯度法估计射影深度,通过奇异值分解实现射影重构.利用共轭梯度法确定Kruppa方程中的未知比例因子,然后利用所确定的比例因子线性求解Kruppa方程.进而标定摄像机内参数.在摄像机内参数已知的情况下,求解一个满足欧氏重构条件的非奇异矩阵,然后通过此矩阵将射影重构变换为欧氏重构.实验结果表明所给出的算法是行之有效的。
In this paper, the general framework of projective reconstruction based on SVD is introduced. Taking the measurement matrix rank 4 as the constraint, the affine projection is used to approximate perspective projection, the projective depths are iteratively estimated by using conjugate gradient method. The projective reconstruction is obtained by SVD of the measurement matrix. The conjugate gradient method is used to estimate the unknown scale factors in Kruppa's equations, then use the scale factors to solve Kruppa's equations linearly and calibrate the camera intrinsic parameters. In the case of the intrinsic parameters of camera are known, a non-singular matrix is evaluated, which satisfies the conditions of euclidean reconstruction, this matrix can transform projective reconstruction into euclidean reconstruction. The result indicates the algorithm is efficient.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2003年第4期407-411,共5页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金(No.60143003)
安徽省自然科学基金(No.01042206)
