基于Alexnet的金相识别研究.docxVIP

MECHANICALENGINEERMECH

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

机械工程师

MECHANICALENGINEER

MECHANICALENGINEERMECHANICALENGINEERMECHANICALENGINEERMECHANICALENGINEERMECHANICALENGINEERMECHANICALENGINEERMECHANICAL

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

MECHANICALENGINEER

机械工程师

MECHANICALENGINEER

PAGE38圆园

PAGE38

圆园21年第9期网址：电邮：hrbengineer@163.com

39网址：

39

网址：电邮：hrbengineer@163.com圆园21年第9期

基于Alexnet的金相识别研究

谭兆湛，官振林

（华南理工大学广州学院，广州510800）

摘要：对《机械工程材料综合实验》课程后期的金相组织图像处理，以深度学习代替人眼观察进行自动识别分类。通过建立金相组织图像库，分析各类材料金相组织的不同形态特征，结合Alexnet神经网络模型，研究面向不同组织金相识别的

Alexnet迁移学习方法,实现金相图像的自动识别。在4类不同的金相组织图像样本库上进行图像识别实验，准确率达到99.63%，能够快速地实现不同材料的区分。与通过人眼观察的方法对比，可大大减少人工操作的工作量，且方法迅速准确。关键词：金相组织；图像识别；深度学习；Alexnet

中图分类号:TP391.7 文献标志码:粤 文章编号：员园园圆原圆猿猿猿（圆园21）09原园038原园4

ResearchonMetallographicIdentificationBasedonAlexnet

TANZhaozhan, GUANZhenlin

(GuangzhouCollegeofSouthChinaUniversityofTechnology,Guangzhou510800,China)

Abstract:Automaticrecognitionandclassificationbasedondeeplearninginsteadofhumaneyeobservationinthemetallographicstructureimageprocessingofcomprehensiveexperimentofmechanicalengineeringmaterialscourse.Throughtheestablishmentofmetallographicstructureimagedatabase,thedifferentmorphologicalcharacteristicsofvariousmaterialsmetallographicstructuresareanalyzed.CombinedwiththeAlexnetn