摘要
传统的图像校正方法需要在拍摄现场布放若干标定点。本文给出了一种新的算法,根据摄像机内部的物理结构,结合DLT(直接线性变换)法等相关知识,构造四个虚拟标定点,根据这些虚拟标定点求出校正系数。该算法无需在拍摄现场安放任何标定点就能够完成对图像的校正,得到物体的真实形状。文中使用该算法对六幅拍摄图像进行了校正,检验了该算法的有效性,并分析了产生微小误差的原因,给出了减小误差的方法。
Some calibration points need to be placed in the shooting region by the traditional image correction, and then the image is corrected based on the knowledge of the computer vision. A novel algorithm without a calibration point is proposed for image correction. Four virtual calibration points are constructed with the help of some information on camera＇s internal physical structure and the Direct Linear Transformation （DLT） algorithm, and the correction coefficient is calculated by means of these virtual calibration points. According to the correction coefficient, the true shape of an object is obtained. The new algorithm is tested by correcting six images. Furthermore, the reasons for a micro-error are analyzed. Meanwhile, some methods for decreasing error are presented.
出处
《光电工程》
CAS
CSCD
北大核心
2009年第1期31-35,共5页
Opto-Electronic Engineering
基金
国家人事部留学回国人员科技活动项目择优资助课题
关键词
计算机视觉
图像校正
DLT
标定点
computer vision
image correction
DLT
calibration point
