基于图像传感器与机器学习的划片机对准算法  

Alignment Algorithm of Wafer Dicing Saw Based on Image Sensor and Machine Learning

作　　者：汪忠伟 张紫薇 WANG Zhong-wei;ZHANG Zi-wei(Shenyang Academy of Instrumentation Science Co.,Ltd.,Shenyang 110043,China)

机构地区：[1]沈阳仪表科学研究院有限公司,辽宁沈阳110043

出　　处：《仪表技术与传感器》2021年第7期48-53,共6页Instrument Technique and Sensor

摘　　要：提出了一种使用机器学习的方式实现激光划片机中晶圆精确对准的新算法。算法采用图像传感器采集晶圆图像,经过深度神经网络学习,建立机器学习模型。借助模型泛化过程实现晶圆粗对准。在粗对准基础上,进一步采集晶圆图像,并经过特征提取和拟合形成回归模型,使用该模型实现晶圆细对准。在紫外激光划片机平台上进行了实验,经过粗对准和细对准两步,实现了晶圆的精确对准,取得了良好的效果。A new algorithm for precise alignment in a laser wafer dicing saw using machine learning was proposed.The algorithm used image sensors to collect wafer images,and then through deep neural networks,a machine learning model was established.Coarse wafer alignment was realized with the help of model generalization process.On the basis of coarse alignment,the wafer image was further collected,and a regression model was formed through feature extraction and fitting,and the model was used to achieve fine wafer alignment.Experiments were carried out on the platform of a UV laser wafer dicing saw.After two steps of coarse alignment and fine alignment,precise alignment of the wafer was achieved and good results were achieved.

关 键 词：划片机 机器学习 对准 深度学习 回归模型 晶圆 

分 类 号：TH3[机械工程—机械制造及自动化]