一种基于等腰梯形的摄像机自标定方法

Method of camera self-calibration based on isosceles trapezoid

给出一种基于等腰梯形的摄像机自标定方法.给定k(k≥6)幅包含等腰梯形的图像,分别检测梯形的 4条边,计算梯形平行边方向上的消影点和 4个角点.根据射影几何中的调和共轭理论,确定梯形 2条平行边的中点.利用正交方向上的消影点对绝对二次曲线C的约束,建立 (k-1)个关于C的方程,对解C进行Cholesky分解,从而获得摄像机内参数.实验结果表明,所提方法具有较高的定标精度.该方法不涉及图像匹配,无需等腰梯形的大小和位置等几何信息,原理简单.

To develop this method, k(k ≥ 6) images containing isosceles trapezoids are given. The four edges of each isosceles trapezoid are detected and the vanishing point of the direction of parallel edges and four corner points are worked out. By the theory of harmonic conjugate in projective geometry, midpoints of two parallel edges are determined. Using the constraints of vanishing points on orthogonal directions to absolute conic C, (k-1) equations on C can be established. Applying Cholesky decomposition to the solution C, the camera intrinsic parameters can be obtained then. Experimental results show that the proposed method has a quite high accuracy of calibration. Image correspondence is not involved in this approach, and the geometric information of isosceles trapezoids such as size and position etc. are not needed. There it is of simple principle.