外文翻译原文-基于机器视觉的苹果果实识别_第1页
外文翻译原文-基于机器视觉的苹果果实识别_第2页
外文翻译原文-基于机器视觉的苹果果实识别_第3页
全文预览已结束

下载本文档

版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领

文档简介

AdvanceJournalofFoodScienceandTechnology2(6):325-327,2010ISSN:2042-4876MaxwellScientificOrganization,2010Submitteddate:October22,2010Accepteddate:November13,2010Publisheddate:November30,2010CorrespondingAuthor:M.B.Lak,DepartmentofAgriculturalMechanization,ScienceandResearchBranch,IslamicAzadUniversity,Tehran,Iran325AppleFruitsRecognitionUnderNaturalLuminanceUsingMachineVision1M.B.Lak,2S.Minaei,3J.Amiriparianand2B.Beheshti1DepartmentofAgriculturalMechanization,2DepartmentofAgriculturalMachineryEngineering,ScienceandResearchBranch,IslamicAzadUniversity,Tehran,Iran3DepartmentofMechanicsofAgriculturalMachinery,FacultyofAgriculture,Bu-AliSinaUniversity,Hamedan,IranAbstract:Inthisstudy,edgedetectionandcombinationofcolorandshapeanalyseswasutilizedtosegmentimagesofredapplesobtainedundernaturallighting.Thirtyimageswereacquiredfromanorchardinordertofindanappleineachimageandtodetermineitslocation.Twoalgorithms(edgedetection-basedandcolor-shapebased)weredevelopedtoprocesstheimages.Theywerefiltered,convertedtobinaryimages,andnoise-reduced.Edgedetectionbasedalgorithmwasnotsuccessful,whilecolor-shapebasedalgorithmcoulddetectapplefruitsin83.33%ofimages.Keywords:Appleharvesting,color-shapebasedalgorithm,edgedetection,machinevisionINTRODUCTIONFreshfruitsharvestingisasensitiveoperation.Itsprofitabilitymaybeinfluencedbylaborinaptitude,costsandunavailability,lowqualityharvesting,andoperationuntimeliness.So,mechanizedharvestingoperationmaysolvetheproblems.MechanizationofapplefruitsharvestingincountrieslikeIranthatisthe4thappleproducerintheworld(FAO,2009)isanessentialneed.Mechanizedfruitharvestingmaybemechanicallyorautomatically.Problemsaccompanyingwithmechanicalharvestingresultedindevelopmentofroboticharvestingmethods,therebyprototypemachinevisionbasedharvestershasbeenincreasinglybeingdeveloped.ParrishandGoksel(1977)andBulanonandKataoka(2010)studiedroboticapplefruitharvesting.Theautomatedharvestingsystemshouldperformthefollowingoperations:(1)recognizeandlocatethefruit;(2)reachforthefruit;(3)detachthefruitwithoutcausingdamagebothtothefruitandthetree;and(4)moveeasilyintheorchard(Sarig,1990).Thefirstoperationneedsdevelopmentofappropriatemethodstodetectandlocatethefruits.Usingphotometricinformationbased(SchertzandBrown,1968)andinfraredlaserrangefinding(Jimenezetal.,2000)methodsweredeveloped.While,imageprocessingbasedmethodshavebeenusedtodetectandlocatedthefruits(BulanonandKataoka,2010;Satish,2007;Harreletal.,1989).Bothintensity/colorpixel-basedandshape-basedanalysismethodswereappropriatestrategiesfortherecognitionoffruits,butsomeproblemsarosefromthevariabilityofthesensedimageitselfwhenusingCCDcameras,whichareverysensitivetochangesinsunlightintensityaswellasshadowsproducedbytheleaves(Jimenezetal.,2000).SincenoresearchhasbeenreportedonroboticappleharvestinginIran,thispaperfocusesonrecognitionofapples,asthefirststageofappleroboticharvesting.Recognitionofapplefruitsusingmachinevisionundernaturaldaylightconditionswastheobjectiveofthisstudy.ThirtyimagesofRedDeliciousapplecanopywereselectedrandomlyfromphotostakenofappletreesinautumn.TheimagesweretakenfromHamedangroves,inIran.MATERIALSANDMETHODSThirtydigitalimageswereobtainedunderuncontrolleddaylightconditions.Imageframeswere30722304pixelsintheJPEGformat.Adigitalcamera(Sony,DSC-H5,ColorCCDCamera)wasusedtoacquiretheRGBimages.Imageprocessingalgorithm:Thegoalwasfindinganappleineachimageobtainedinuncontrolledlightingconditions.Inordertosegmenttheacquiredimages,twoalgorithmsweredeveloped:edgedetectionbasedandcolor-shapebased.Edgedetection-basedalgorithm:Initialconsiderationsshowedthatthegreengray-scaleimageincludedmostofAdv.J.FoodSci.Technol.,2(6):325-327,2010326thedesiredobjects.Canny(1986)methodwasusedtodeterminetheedgesofapplesingreengray-scaleimage(Fig.1b).Color-shapebasedalgorithm:Thealgorithmwasimplementedbasedonthefollowingsteps:CTheimageswerefirstenhanced.AGaussianlow-passfilterwasusedtoreducethenoiseasmuchaspossible.NoiseportendstounequalcolorintensitydistributionintheoriginalimagesthatformedshadesandshinyregionsintheimagesTheGaussianfilterwasa250250pixelmatrixwithstandarddeviationsof200,whichlimitimagefrequenciestolessthan200MegaHertz(MHz).Filteredimageswerenoise-reducedbyremovinghighfrequencies(morethan200MHz).Filteringtheimagecausedblurringwhichnoisewasreduced(Fig.2a)CFilteredimageswerethenconvertedtobinaryforminordertobeprocessed(Fig.2b)CBinaryimageswereprocessedtoreducetheexistingnoiseafterconvertingimages.Inthisstage,noisewasdefinedastheareasdetectedasfeaturesotherthanapples.Thisstageoftheprojectisshape-basedprocessingofcolor-basedprocessedimages(Fig.2c)CBinary,noise-removedimageswerelabeledtoextracttheapple(Fig.2d)(a)(b)Fig.1:Edgedetectionbasedalgorithm.a)Originalimage,b)Edge-detectedimage(a)(b)(c)(d)Fig.2:Color-shapebasedalgorithm.a)Filteredimage,b)Binaryimage,c)Noise-reducedbinaryimage,d)LabeledimageAdv.J.FoodSci.Technol.,2(6):325-327,2010327RESULTSANDDISCUSSIONThemainideawastodevelopageneralalgorithmundervariousnaturallightingconditions.Thereby,nosupplementallightingsourcewasusedtocontroltheluminance.Sincetheimageswereacquiredunderuncontrollednaturaldaylightconditions,theyincludedtreecanopiesincludingtreebranches,leaves,fruits,sky,etc.Eachobjectoftheimagehasitsownedges,makingimagesetsofedgesofwhichtheappleisjustasubset.So,edgedetectionalgorithmwasnotsuccessful(Fig.1).Color-shapebasedalgorithmdetectedtheimageobjectsintheimagesbetter;however,itwasmorecomplicatedthantheedgedetection.Thestageofcolorprocessingblurredtheimageanditsoutputwasanimagewithlowcontrasthavingdistributedcolors(Fig.2a).Thus,theimagewasnoise-reducedinwhichthenumberofobjectswerelessthanthattheoriginalimage.Convertingtheimagetobinaryformandshape-basedanalysismadethenoiseaslowaspossible(Fig.2b,c).Color-shapebasedalgorithmwasabletodetecttheapplesin25of30images.Inotherwords,theaccuracyofthealgorithmwas83.33%.Figure2showstheprocedureofcolor-shapebasedalgorithm.CONCLUSIONInthisstudy,twoalgorithmsweredevelopedandcomparedtodetectoneappleineachimage.Nolightingcontrolwasexercisedtostandardizeluminanceoftheacquiredimages.However,edgedetectionbasedmethodwasnotsuccessful;color-shapebasedalgorithmwasabletodetectapplesin83.33%ofimages.REFERENCESBulanon,D.M.andT.Kataoka,2010.FruitdetectionsystemandanendeffectorforroboticharvestingofFujiapples.Agric.Eng.Int.CIGRJ.,12(1):203-210.Canny,J.,1986.Acomputationalapproachtoedgedetection.IEEET.PatternAnal.,8(6):679-698.FAO,2009.FoodandAgricultureOrganizationOfficialWe

温馨提示

  • 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
  • 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
  • 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
  • 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
  • 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
  • 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
  • 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。

评论

0/150

提交评论