基于双曲正切的亚像素边缘检测方法及评价  

A Sub-pixel Edge Detection Algorithm and Evaluation based on Hyperbolic Tangent

作　　者：韦光[1] 谭庆昌[1] 刘国栋[1] 熊磊[1] 杨鑫[1] 刘健英[1] 

机构地区：[1]吉林大学机械科学与工程学院,吉林长春130022

出　　处：《工程与试验》2010年第3期1-4,77,共5页Engineering and Test

摘　　要：精确定位图像边缘是保证视觉测量精度的基础。根据物体图像边缘的灰度分布特点,本文提出一种亚像素边缘检测方法,采用双曲正切函数对图像边缘的灰度分布进行拟合,通过最小二乘法得到边缘的亚像素位置。对提出的边缘检测方法的精度进行了评价,以高精度棋盘格的尺寸测量误差作为准则,采用实际边缘并结合摄像机标定技术,考虑了镜头畸变和透视失真等因素。试验表明,提出的双曲正切法检测亚像素边缘相比空间矩法和梯度插值法具有更高的精度,同时能够满足机械工业中的实时测量需求。Locating edges in images accurately is a basic to ensure the precision of measurement.According to the characteristics of edge intensity distribution in actual images,this paper puts forward a method to detect sub-pixel edges,which fits the edge intensity distribution with hyperbolic tangent function,and gets subpixel edges through least-square method.Moreover,this paper makes an evaluation of accuracy of the algorithm proposed,taking measurement error of high-accuracy checkerboard dimensions as criterion.This evaluation method combines with camera calibration technique,which considers the factors of lens distortion and perspective distortion.Experiments show that hyperbolic tangent algorithm proposed has higher precision than spatial moment-based algorithm and gradient interpolation algorithm,and can meet the requirement of real-time measurement in mechanical industry.

关 键 词：双曲正切 亚像素 边缘检测 评价 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]