Abstract: This paper proposed a new camera self-calibration method based on image sequences, which combined the theory of photogrammetry and computer vision. Firstly, the camera foundation matrix F was obtained by making use of the 8 points algorithm according to the image matching points, and the matrix C was got based on utilizing Kruppa equation and the foundation matrix F, then the matrix C was decomposed with Cholesky method to get the camera internal parameter matrix K. Secondly, the relative orientation and absolute orientation were carried on taking the internal parameters as the initial value, then the image matching points's 3d space coordinate was got by least squares forward intersection. Finally, the accurate camera internal parameters, external parameters and distortion coefficients were calculated by three-dimensional direct linear transformation and bundle adjustment methods. This self-calibration method does not depend on the particular geometric constraints in the scene, and it can be carried out as long as there are matching points, so it has broad applicability. Both simulation data and real images were used to test the method, and the results show that: this new method gets high precision and low computational complexity; it is a new and valid method for camera self-calibration.

Key words: camera self-calibration, foundation matrix, Kruppa equation, relative orientation, absolute orientation, bundle adjustment

摘要： 提出了一种新的结合摄影测量和计算机视觉相关理论的摄像机自标定方法。首先通过序列图像的匹配点对，利用计算机视觉理论中的8点法求得摄像机基础矩阵F，通过矩阵F利用Kruppa方程求得矩阵C，对矩阵C进行Cholesky分解得到摄像机的内参数矩阵K，然后将求出的内参数作为初始值，利用摄影测量理论进行相对定向和绝对定向，最小二乘前方交会计算得到匹配点对的三维空间坐标，最后由匹配点对的三维空间坐标及其图像坐标，采用三维直接线性变换和光束法平差方法解算出摄像机内、外参数及畸变系数。该方法不依赖于特定的场景几何约束条件，只要序列图像之间有匹配点对，就可以进行自标定工作，具有广泛的适用性。模拟数据和真实图像的实验结果表明：该方法计算过程简单，标定精度高，是一种值得借鉴的摄像机自标定方法。

关键词: 摄像机自标定, 基础矩阵, Kruppa方程, 相对定向, 绝对定向, 光束法平差