基于图像的昆虫远程自动识别系统的研究

1882008(11000942100050)Internet7Hu121640202096%YangHongzhen,ZhangJianwei,LiXiangtao,etal.Remoteautomaticidentificationsystembasedoninsectimage[J].TransactionsoftheCSAE,2008,24(1)188192.(inChinesewithEnglishabstract)2007-03-192007-05-172004B33“”2006BAD08A011972126100050Email:[e-mail]19442100094Email:[e-mail][1-3]241Vol.24No.118820081TransactionsoftheCSAEJan.20081189javaphpaspC++C++socket1Fig.1Sketchmapofremotetransmissionofinsectimagesandidentificationresults127Hu[45](1)7Hu(2)(13)RALRoundnessR24LARπ=1Rxyfxy,p+q∑∑=xyqppqyxfyxm),(2R∑∑−−=xyqppqyxfyyxx),()()(µ3yx,0010/mmx=40001/mmy=5Rrpqpq00µµη=12++=qpr,p+q=2,3,6Hu,1962702201ηηφ+=7()2112022024ηηηφ+−=8()()2032121230333ηηηηφ−+−=91902008()()20321212304ηηηηφ++−=10()()()()[]()()()()[]20321212300321032120321212301230123053333µηηηηηηηηηηηηηηηφ+−++−++−++−=11()()()[]()()03211230112032121230022064ηηηηηηηηηηηφ++++−+−=12()()()()[]()()()()[]20321212300321031220321212301230032173333µηηηηηηηηηηηηηηηφ+−++−++−++−=13RGBLLLL=2),,(Max),,(MinBGRBGR+14RGBL255321(1/4,1/2,1/4)RGBL(R)(G)(B)(15)R=bgrr++15G=bgrg++16B=bgrb++17RGRG1633[(1/4,1/2,1/4)T(1/4,1/2,1/4)]Fig.2Flowchartofautomaticidentificationofinsects(16)Xi={x1,x2,x3,x4,x5,x6,x7,x8,x9,,xn,IDi}18IDiIDiXCC=⎥⎥⎥⎥⎦⎤⎢⎢⎢⎢⎣⎡………nnnnnnXXXXXXXXX212222111211####19RBF11913Fig.3Combinedidentificationofclassifyingmachineofdifferentcharacteristics16202012RGBLRGBLRGBL0.3RGBL0.3596.2%16GaeanamaculataHuechyssangninea(Lycormadelicatula)(Metonymiaglandulosa)(Erthesinafullo)(Poecilocorislewisi)Anomalacupripes(Anomalacorpulenta)Aspidomorphamiliaris(Craspedontaleayana)(Dorcustitanus)(Teleogryllusemma)(Pseudophyllustitan)(Papilioparis)(Papilioalcmenor)Kallimainachus13[6-12]1[12]2[10]RGBHIS3RGBL119220082HuHu[8][][1].[J].200239(4):315316.[2].[J].200418(6):337338.[3].[J].200532(3):148150.[4]HuMK.Visualpatternrecognitionbymomentinvariants[J].IEEETransactionsonInformationTheory,1962,8:179187.[5].[D].20061475.[6].[J].200723(5)154158.[7].[J].200016(3)48.[8].[J].200420(5):4346.[9].[J].2004,20(3):109112.[10].[J].2000,16(3):115117.[11].[J].200222(4):410414.[12].[J].200633(7):96100.RemoteautomaticidentificationsystembasedoninsectimageYangHongzhen1,2,ZhangJianwei1,LiXiangtao2,ShenZuorui1(1.CollegeofAgronomyandBiotechnolog,ChinaAgriculturalUniversity,Beijing100094,China;2.BeijingMuseumofNaturalHistory,Beijing100050,China)Abstract:Onlyafterthepestwasidentified,couldthepurposefulcontrollingofthepestbeconductedinagriculturalproduction.Theidentificationofinsectsisoneofthemostimportantjobsinentomology.Anovelmethodbasedoncomputervisiontechnologywitharemotewaycanmakeavaluablecontributiontoreducingtheburdenofroutineidentifications.Itcanalsoprovideapotentialsolutiontothegrowingburdenofroutinespeciesidentificationspresentlyfacedbyadwindlingcommunityofexperttaxonomists.Insectimagewascapturedanduploadedtoidentificationserverfrominternet.Identificationprocesswasdonebyserverautomaticallyandresultsweregivenbacktouserimmediately.ShapeandcolorcharacteristicswereusedtoformtheseparatedpatternvectorandRBFnetworkwasusedasclassifierforeachpattern.Eachclassifierwaslinearlycombinedtogivethefinalidentificationresult.Shapepatternconsistsofrectangularity,elongation,roundness,sphericity,lobationandsevenmovementinvariants.R,G,B,LhistogramandR.Gchromaticityhistogramweretreatedasseparatedcolorcharacteristicvector.Theidentificationsystemwasusedtotest16species,40samplesforeach,with20ofthe40samplesfor“training”andtheremainderfor“test”.Theaccuracyreachesabove96%.Keywords:insect;automaticidentification;remoteidentification;imageprocessing;computervision