基于畸变模型优化的网络摄像机标定*

针对低分辨率网络数码摄像机用于定量量测的标定问题, 提出一种非线性畸变的构建与优化方法。该方法结合多项式模型与传统的畸变模型；利用回归分析原理对模型进行自动精化，以优选显著的畸变参数，从而建立了一种自适应的摄像机畸变模型。运用构建的畸变模型，利用自检校光束法平差方法对同种型号的三个低成本网络摄像机实施标定和畸变纠正。结果表明，该模型有效补偿了通用畸变模型残存的畸变系统差，同时有效的避免了过度参数化的问题，有助于提高摄像机的标定精度和稳健性，使得摄像机的标定精度达到子像素。

Web-Camera calibration with optimum distortion model

To solve calibration problem for low-cost web-cameras used in quantitative measurement, this paper proposed a method to develop the optimum distortion model. The approach includes deriving an extended model from the combination of conventional model with polynomial model ,and refining model based on regression analysis principle. Based on the refined distortion model, three low-cost web-cameras were calibrated with self-calibration bundle adjustment. Experimental results show that it can attain sub-pixel accuracy in images, which is superior to one pixel level from traditional distortion model or the combined model with all distortion parameters. It is also concluded the optimum model can compensate the distortion systematic residuals and improve the robustness for self-calibration bundled adjustment due to avoiding over-parametric problem.