基于消隐点无穷单应的摄像机焦距精确自标定方法

An Accurate and Real-Time Focal-Length Self-Calibration Method Based on Infinite Homography between Vanish Points

利用一组平行线在不同姿态摄像机图像平面中对应消隐点间的无穷单应关系和摄像机相对姿态信息,提出了一种摄像机焦距的高精度实时标定方法。该方法仅通过摄像机在任意两个位置下拍摄同一组空间平行线,基于消隐点对之间的无穷单应关系构建约束,求解焦距参数;将对应光心与消隐点连线的平行程度作为优化指标,利用Nelder-Mead非线性单纯型法实现焦距参数的优化,有效地抑制了图像噪声和姿态测量误差,提高了标定结果的精度和算法的稳健性;大量仿真结果验证了该算法可以在任意相对姿态下实现,且具有精度高、稳健性强、实时解算的优点。

Based on the infinite homography relation between the vanish points on images captured in different camera positions and the relative position information, a self-calibration method is proposed to calibrate the focal- length of camera accurately. By projecting parallel lines in two positions, this method achieves a pair of vanish points in image plane, an infinite homography constraint is established and the focal-length is solved. Considering the parallel extent of lines that connecting corresponding optical centers and vanish points as the optimization index of Nelder-Mead nonlinear simplex method, focal-length optimization is achieved, the effect of image noises and rotation uncertainties are restrained effectively, and the accuracy of calibration result and algorithm robustness are improved greatly. Plenty of simulations show that this method proposed is both accurate, robust and real-timed.