摘要
Abstract
With the outbreak of the artificial intelligence industry,the demand for AI chips will have a significant growth rate in the next five years.In the process of producing wafer chips,in order to improve the production capacity and yield control of wafer chips,machine vision technology introduced is introduced to study the defect detection of wafer chips and proposes a defect detection method based on Halcon.It includes two major parts:hardware and software.The hardware part includes industrial cameras and light sources.The software part includes Halcon camera calibration,Fourier transform from spatial domain to frequency domain,defect extraction,threshold segmentation,connected domain analysis,noise filtering,and calculation results.Through this method,two possible types of defects(cracks and dirt)on the surface of the chip are tested on sample chips.The experimental results show that the accuracy of defect detection for cracks is 98%,and the accuracy of defect detection for dirt is 97%.At the same time,this method has a fast detection response speed,high stability,and strong anti-interference ability against changes in the chip surface and light interference.It is a feasible method for detecting defects in wafer chips.关键词
机器视觉/Halcon/缺陷检测/晶圆芯片Key words
machine vision/Halcon/defect detection/wafer chip分类
信息技术与安全科学
