基于机器视觉的图像识别和瑕疵检测研究.pdfVIP

informationoftheworkpiececenterpointisconveyedtosuckertoachieve

palletizing.

(3)Inregardtotheshortcomingincylindersurfacedefectdetection,thedefect

detectionmethodbasedonmachinevisionanddeeplearningisanalyzed.Adefect

detectionsolutionthatcombinesmachinevisiontechnologyanddeeplearning

technologyisproposed.Inthisway,whenthedefectdetectionrateisinsufficient

throughtraditionalapproaches,thismethodcandefecttheimageforthesecondtime

viathesupportivedeeplearningsoastoenhancethedefectdetectionrate.Inaddition,

experimentsaredesignedtoverifytheeffectivenessofthismethod.

(4)BasedonMFC(MicrosoftFoundationClassLibrary),anapplicationprogram

foridentification,localizationanddefectdetectionisdevelopedbyintegratingthe

abovealgorithmsintodifferentfunctionalmodules.Thepracticalapplicationofthis

softwareprovesthatwithconvenientoperationandstableprocedurerunning,the

workpieceidentificationandlocalizationfunctioncanacquiretheworkpieceposition

information,andthecombineddefectdetectionmethodcanobviouslyimprovethe

rateofdetectingdefectivecylinders.Therefore,theprojectexpectedeffectis

achieved.

Keywords：MachineVision，MultipleIdentificationandLocalization，Defect

Detection，DeepLearning，ConvolutionalNeuralNetwork

目录

摘要I

ABSTRACTII

1绪论1

1.1研究背景与意义1

1.2国内外机器视觉识别与瑕疵检测研究与发展2

1.2.1机器视觉2

1.2.2国内外机器视觉识别发展概况3

1.2.3国内外机器视觉瑕疵检测发展概况5

1.3本文研究内容及组织结构6

1.3.1研究内容6

1.3.2结构安排7

2机器视觉系统的技术基础9

2.1硬件结构9

2.1.1视觉系统分类9

2.1.2光源类型9

2.1.3图像采集卡10

2.2瑕