面向刀具磨损图像区域分割的改进分水岭算法  

An Improved Watershed Algorithm for Segmenting Tool Wear Images

作　　者：刘建军 刘丽冰[1] 彭伟尧 张艳蕊 杨泽青[1] Liu Jianjun;Liu Libing;Peng Weiyao;Zhang Yanrui;Yang Zeqing(School of Mechanical Engineering,Hebei University of Technology,Tianjin 300130,China;Experimental Training Center,Hebei University of Technology,Tianjin 300401,China)

机构地区：[1]河北工业大学机械工程学院,天津300130 [2]河北工业大学实验实训中心,天津300401

出　　处：《机械科学与技术》2020年第5期729-735,共7页Mechanical Science and Technology for Aerospace Engineering

基　　金：国家自然科学基金项目(51305124);河北省科技计划项目(16211803D);天津市自然科学基金项目(16JCYBJC19100);河北省自然科学基金项目(E2017202294)资助。

摘　　要：原始刀具图像通常存在背景纹理复杂、噪声大等问题,导致磨损区域分割结果的准确性较差,为此本文提出了一种基于形态学成分分析(MCA)的改进分水岭算法,用于提取刀具磨损区域并估算其面积。首先分析了刀具磨损图像各组成成分的形态差异;然后研究了各成分对应字典的选取方法,将原始刀具图像分解成目标刀具图像、背景图像和噪声;最后对目标刀具图像使用分水岭算法提取磨损区域并估算面积。以铣刀磨损图像作为样本完成了多次方法验证,结果表明:传统分水岭算法的检测误差为80%左右,而该方法的检测误差为5%以下,可见使用该算法可以分割得到更加准确的磨损区域。An original tool image usually has complex background texture and high noise,which lead to the poor accuracy of wear region segmentation.Therefore,an improved watershed algorithm based on morphological component analysis(MCA)is proposed to extract the tool wear region and estimate its area.Firstly,the morphological differences of the components of the tool wear image are analyzed.Then,the method for selecting the corresponding dictionary of each component is studied and used to decompose the original tool image into the target tool image,background image and noise.Finally,the watershed algorithm is used to extract the wear region of the target tool image and to estimate its area.Milling wear images are used as samples to perform a number of validations of the algorithm.The validation results show that the detection error of the traditional watershed algorithm is about 80%,while the detection error of our algorithm is less than 5%,so it is concluded that our algorithm can segment wear regions more accurately.

关 键 词：刀具磨损 图像分割 形态学成分分析 分水岭算法 

分 类 号：TH166[机械工程—机械制造及自动化]