正则化方法在鱼眼相机检校中的应用  

Regularization Method in the Fish-Eye Camera Calibration

作　　者：李崇辉[1] 原玉磊[2] 郑勇[1] 张超[1] 

机构地区：[1]信息工程大学,河南郑州450001 [2]国防科学技术大学计算机学院,湖南长沙410073

出　　处：《信息工程大学学报》2013年第5期575-579,共5页Journal of Information Engineering University

基　　金：国家自然科学基金资助项目(41174025)

摘　　要：由于能对大范围内的目标同时成像,鱼眼相机已经在全方位视觉和摄影测量等领域得到了广泛的应用。使用鱼眼相机之前需要对其进行检校,然而鱼眼相机参数之间存在较强的复共线性,致使参数解容易受到观测误差的影响。为了解决这一问题,提出了利用正则化方法来进行鱼眼相机检校,阐述了Tikhonov正则化方法的基本原理,介绍了L曲线法确定正则化参数的原理,最后利用实验数据对最小二乘解和正则化解的准确度和精度进行了比较。实验结果表明正则化方法能够有效地减弱检校方程的病态性,能够显著提高鱼眼相机参数解的稳定性。Fish-eye camera has been widely used in Omni-directional vision, photogrammetry and other fields for its simultaneous wide-range-field imaging. It should be calibrated before being used, but strong muhi-colinearity exists between the camera parameters, so parameter solutions are vulnerably affected by observation error. In order to solve this problem, the regularization method is proposed in fish-eye camera calibration. The basic principle of Tikhonov regularization method is expounded, and the L-curve method is introduced to determine the regularization parameters, finally solutions of least square method and regularization method are compared in terms of accuracy and precision. Experiment results indicate that the proposed regularization method can effectively weaken the ill-posed problem of calibration equation, so it can significantly improve the stability of fish-eye camera parameters.

关 键 词：鱼眼相机 检校 TIKHONOV正则化 L曲线 

分 类 号：TB22[天文地球—大地测量学与测量工程]