卷积神经网络的高精密光学元件表面缺陷检测  

Detection of surface defects of high precision optical components based on convolution neural network

作　　者：赵博 史迎馨[1] ZHAO Bo;SHI Yingxin(Tonghua Normal University,Tonghua Jilin 143001,China)

机构地区：[1]通化师范学院,吉林通化143001

出　　处：《激光杂志》2021年第11期185-189,共5页Laser Journal

基　　金：吉林省教育厅科研项目(No.JGJX2019D301)。

摘　　要：为解决高精密光学元件表面缺陷检测方法存在的精度低、耗时长等的缺陷,提出基于卷积神经网络的高精密光学元件表面缺陷智能检测方法。首先分析当前高精密光学元件表面缺陷检测的研究进展,找到引起检测结果不足的因素,然后采集高精密光学元件表面缺陷图像,提取检测特征,并引入卷积神经网络建立缺陷检测分类器,实现高精密光学元件表面缺陷检测。实验结果表明,所提方法的缺陷检测精度超过93%,缩短了缺陷检测时间,平均单次检测时间降低0.7 s以上。In order to solve the defects of low precision and long time-consuming in the surface defect detection of high-precision optical components,an intelligent detection method based on convolution neural network is proposed.Firstly,the research progress of surface defect detection of high-precision optical components is analyzed,and the factors causing the insufficient detection results are found.Then,the image of surface defects of high-precision optical components is collected,and the detection features are extracted.Finally,the convolution neural network is introduced to establish the defect detection classifier to realize the surface defect detection of high-precision optical components.The experimental results show that the defect detection accuracy of the proposed method is over 93%,the defect detection time is shortened,and the average single detection time is reduced by more than 0.7 s.

关 键 词：高精度光学元件 表面缺陷 卷积神经网络 分类器 检测特征 

分 类 号：TN249[电子电信—物理电子学]