Abstract: By constructing the non-linear model，classical digital camera calibration method is reinforced.And the problem of digital camera exact calibration solved depends on fewer sample size data.First，direct linear transform model is used to get the projection matrix.Then inside and outside parameters of digital camera are decomposed by the restraint condition[m3=1].These estimates are fed to the nonlinear optimizer as initial guess for the calibration parameters.The optimal solutions of objective function J=∑pi-p(A,k1,k2,R,ti,Pi)²are solved，by the restraint conditions h^T(A^-1)^TA^-₁h₂=0 and h^T1(A^-1)TA-1h₁=h^T2(A^-1)^TA^-1h₂Consequently，digital camera exact calibration is implemented.The result of single sample data indicates that a high accuracy is obtained by the algorithm presented in this paper.

Key words: camera calibration, non-linear optimal model, stereovision

摘要： 通过构造非线性模型，对经典标定方法进行了补充，同时也解决了在样本数据较少情况下的数码相机精确定位问题。利用线性转换模型得到投影矩阵，设定m₃=1为约束条件，通过分解参数矩阵求出数码相机内外部参数，并将其作为非线性优化模型的初始值；然后构造非线性优化模型，建立目标函数J=∑pi-p(A,k1,k2,R,ti,Pi)²，求解在约束条件h^T(A^-1)^TA^-₁h₂=0和h^T1(A^-1)TA-1h₁=h^T2(A^-1)^TA^-1h₂下的最优解，从而实现数码相机的精确标定。对给出的单样本情况进行了求解，结果表明此算法可以达到较高的精度。

关键词: 相机定标, 非线性优化模型, 立体视觉