一种新的基于Kruppa方程的摄像机自标定方法

主要针对传统的基于Kruppa方程的摄像机自标定算法的欠鲁棒性提出了一种新的二步式标定方法．在新标定方法中，首先利用传统的LM优化算法或遗传算法求解出Kruppa方程中通常需要被消去的比例因子，然后再利用线性方法完成对摄像机的标定．大量的仿真和真实图像实验表明，该方法可以大大提高基于Kruppa方程标定算法的鲁棒性及标定精度．

A Novel Camera Self-Calibration Technique Based on the Kruppa Equations

Conceptually speaking, nearly all the self-calibration techniques reported in the literature can be classified as either the absolute conic based ones or the absolute quadric based ones, and all such techniques rely on the so-called Kruppa equation or its variants implicitly or explicitly. The key problem in the Kruppa equations (at least 3 for calibrating a full perspective camera) resides in their associated unknown scale factors that make the calibration problem inherently non-linear, and not robust. In this paper, a two-step camera self-calibration technique is proposed. In the new technique, the unknown scale factors from the Kruppa equations is eliminated as those in the literature. Instead, first the unknown scale factors are determined via a genetic algorithm or LM algorithm. Then the camera intrinsic parameters are computed by a linear method with the scale factors obtained in the first step. Extensive simulations as well as experiments with real images show the new technique can substantially improve the robustness aspect and increase the calibration accuracy.