CCD光学成像系统中亚像素边缘的精确定位  

Precision Location of Sub-Pixel Edge in CCD Optical Imaging Systems

作　　者：陈新禹[1] 胡英[1] 毕丰隆[1] 

机构地区：[1]大连海事大学自动化研究中心,辽宁大连116026

出　　处：《信息与控制》2014年第2期176-180,共5页Information and Control

基　　金：国家科技重大专项资助项目(2009ZX04001-021)

摘　　要：为了实现对图像边缘的精确定位,提出了一种检测亚像素边缘的两步法.首先,使用Canny算子对图像进行运算,提取像素级初始边缘点.然后,在初始边缘点的梯度方向上利用线性插值法计算6个邻点,并从邻点中选择3个合适点求解二次多项式,从而实现边缘点的亚像素级精确定位.通过实验分析了噪声、边缘角度和边缘灰度差等因素对定位精度的影响.实验结果表明:该方法具有较好的噪声抑制能力,对各个方向的检测精度均优于0.08像素,在边缘灰度差很小的情况下仍能得到优于0.07像素的定位精度.该方法具有较强的鲁棒性,较高的定位精度,在视觉测量和模式识别等领域具有一定的应用价值.To precisely locate the edge of an image,a two-step method to detect sub-pixel edge is proposed.First,the initial edge points at pixel level are obtained using a Canny operator.Second,six neighbor points are calculated using linear interpolation in the direction of every initial edge point＇s gradient.Three suitable points are selected for solving the quadratic polynomial,then the sub-pixel edge points are accurately positioned.The impacts of noise,edge angle,and difference of edge intensities on position accuracy are analyzed by experimental methods.Experimental results indicate that the method is fast and highly precise,and has better noise suppression capability.The detection accuracy for each direction is superior at 0.08 pixel,and the precision is better than 0.07 pixel in the case of small step height.The method is robust and accurate,and can satisfy the requirements of applications in the field of visual detection and pattern recognition.

关 键 词：亚像素边缘检测 CANNY算子 梯度 二次多项式 

分 类 号：TP394.01[自动化与计算机技术—计算机应用技术] TH691.9[自动化与计算机技术—计算机科学与技术]