交叠固定多摄像机协同自标定

Collaborative self-calibration for overlapped stationary multiple cameras

针对视场交叠的固定多摄像机监控场合,提出了一种基于射影分解的多摄像机协同自标定方法。首先,根据对极几何知识得到的基础矩阵和极点,求出匹配点对像素坐标组成的矩阵。然后,对该矩阵进行奇异值分解(Singular Value Decomposition,SVD),求出每个摄像机的投影矩阵和像点的空间坐标。利用旋转因子模约束得到新的投影矩阵,进一步分解可得摄像机的内参数矩阵、旋转矩阵和平移矢量。根据分解出来的像点模糊世界坐标与新建立的实际世界坐标系之间的变换关系,求解出所有像点的三维空间坐标。通过与张氏标定法的对比,证明了该方法的可行性。

For the surveillance of overlapped stationary multiple cameras, a new collaborative self--calibration method is proposed for multiple cameras based on projective factorization. First, the matrix consisted by pixel coordinates of the matching points is de- termined by the fundamental matrix and epipole on two views, which are solved by epipolar geometry. This matrix carl process Sin- gular Value Decomposition（ SVD）, thus the projection matrix and the world coordinates of the matching points can be obtained for each camera. A new projection matrix can be obtained according to the rotation factor modulo-constraint condition. Then this ma- trix is decomposed to the intrinsic matrix, rotation matrix and translation vector. According to the transformation between the newly established world coordinate system and the solved ambiguity world coordinate system, the three-dimensional coordinates of all matching points can be obtained. The experimental results have been compared with the results of Zhang＇ s calibration method, which proves the feasibility of the method.