一种综合景物垂直性约束的自定标算法

An algorithm for auto-calibration with scene orthogonality constraints

考虑到建筑物中普遍存在的平面正交性,将其作为相机自定标的新约束,该约束条件和以往基于相机的约束条件相结合可以使得所求解的绝对二次曲面的欧式特性得到满足,所重建的三维景物更符合实际.该方法的有效性在仿真数据和真实图像数据实验中得到验证.

The plane orthogonality, a comon prior knowledge in most buildings, is considered to form a new constraint on autocalibration equations. A corresponding algorithm is proposed to find the absolute quadric which is close to the practical one. The orthogonalities of the 3D reconstructed scene are satisfied. Furthermore, the robustness of the autocalibration algorithm to the camera critical motions is greatly improved. The effectiveness of our approach is evaluated by the experiments with simulated data and real image sequences.