计算机三维视觉测量中相机标定的圆形校正方法  

Circle calibration method for camera used in 3D measurement based on computer vision

作　　者：陈济棠[1] 徐杜[1] 蒋永平[1] 郭斌[1] 黄品松[1] 

机构地区：[1]广东工业大学信息工程学院,广州510006

出　　处：《电子测量技术》2011年第3期95-98,共4页Electronic Measurement Technology

摘　　要：目前相机标定中大部分用harris角点提取特征点,当标定板与相机夹角过大时,将会减少角点附近的灰度信息,从而影响了特征点提取的精度。研究改进椭圆圆心提取法,提出了一种基于惩罚奖励机制的外轮廓提取法,加快了外轮廓的提取。提出一种新的快速的方法获得多于8个方向的梯度。并对高斯拟合提取亚像素点进行加权处理。实验证明,此算法优于opencv中的算法,可稳定的求得相机内参。In most of the calibration method,people use harris method to extract the Feature points.but when the angle between the camera and the board is large,it will reduce the imformation arround the feature points.thus will impact the Accuracy.in this paper,we improve the circle exctract method,proposed a fast method extracting the outline base on award and punish way.Propose a quick method to get more than 8 dirction.Weighted the Gaussian fitting to get stable Sub-pixel point.Experimental results show that using this method we can Stability get the internal parameters capare with opencv calibration method.

关 键 词：圆形标定 高斯拟合 椭圆拟合 亚像素提取 

分 类 号：TN911.73[电子电信—通信与信息系统]