基于离焦图像的晶圆表面缺陷检测  

Wafer Surface Defect Detection Based on Defocused Image

作　　者：何睿清 周鑫宇 张慧 郝飞[2] 宋佳潼 HE Ruiqing;ZHOU Xinyu;ZHANG Hui;HAO Fei;SONG Jiatong(School of Information and Communication Engineering,Nanjing Institute of Technology,Nanjing 211167,China;School of Mechanical Engineering,Nanjing Institute of Technology,Nanjing 211167,China)

机构地区：[1]南京工程学院信息与通信工程学院,南京211167 [2]南京工程学院机械工程学院,南京211167

出　　处：《组合机床与自动化加工技术》2023年第7期110-113,118,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金项目(61905108);江苏省研究生科研与实践创新计划项目(SJCX21_0941,SJCX21_0913);南京工程学院校级科研基金项目(YKJ201868)。

摘　　要：缺陷检测在晶圆生产中起着关键作用,在利用机器视觉检测晶圆缺陷时,无论晶圆有无缺陷均进行自动对焦,从而导致检测效率降低,针对该问题,提出了一种基于离焦图像的晶圆表面缺陷检测方法。可在对焦前下对晶圆进行评估,检测缺陷后进行对焦,首先在频域中对晶圆图像进行自适应背景估计和去除,得到光照均匀、特征突出的检测图像;然后使用区域合并的优化分水岭算法进行图像分割;最后对分割后缺陷区域进行缺陷识别。实验结果表明,该方法可以准确鉴别晶圆图像是否存在缺陷,对于存在缺陷的样本,缺陷检测准确率可达95.3%,提高了检测效率。Defect detection plays a key role in wafer production,however,autofocus is performed regardless of whether the wafer is defective,which could reduce detection efficiency while using machine vision technology for wafer detection.To solve this problem,this article proposes a method for wafer surface defect detection based on defocused images.This method can evaluate the wafer before focusing,and focuses after defect detection in advance.Firstly,adaptive background estimation is performed and removed on the wafer image in the frequency domain,and the detection image with uniform illumination and prominent features is obtained.Then,using the optimized watershed algorithm for region merging to perform image segmentation.Finally,the defect area after segmentation is identified.Experimental results show that the method can accurately identify whether there are defects in the wafer image,and the detection accuracy of this method can reach 95.3%for samples with defects,which improves the detection efficiency.

关 键 词：晶圆缺陷检测 频域图像处理 背景估计 分水岭算法 特征识别 

分 类 号：TH165[机械工程—机械制造及自动化] TG68[金属学及工艺—金属切削加工及机床]