matlab中英文翻译文献.docx

matlab中英文翻译文献.docx

《matlab中英文翻译文献.docx》由会员分享，可在线阅读，更多相关《matlab中英文翻译文献.docx（11页珍藏版）》请在冰豆网上搜索。

matlab中英文翻译文献

附录英文原文

Scenerecognitionforminerescuerobot

localizationbasedonvision

AbstractAnewseenerecognitionsystemwaspresentedbasedonfuzzylogicandhiddenMarkovmodel（HMM）thatcanbeappliedinminerescuerobotlocalizationduringemergencies.Thesystemusesmonocularcameratoacquireomni-directionalimagesofthemineenvironmentwheretherobotlocates.Byadoptingcenter-surrounddiffereneemethod,thesalientlocalimageregionsareextractedfromtheimagesasnaturallandmarks.TheselandmarksareorganizedbyusingHMMtorepresentthescenewheretherobotis,andfuzzylogicstrategyisusedtomatchthesceneandIandmark.Bythisway,thelocalizationproblem,whichisthescenerecognitionprobleminthesystem,canbeconvertedintotheevaluationproblemofHMM.Thecontributionsoftheseskillsmakethesystemhavetheabilitytodealwithchangesinscale,2Drotationandviewpoint.Theresultsofexperimentsalsoprovethatthesystemhashigherratioofrecognitionandlocalizationinbothstaticanddynamicmineenvironments.

Keywordsrobotlocation;scenerecognition;salientimage;matchingstrategy;fuzzylogic;hiddenMarkovmodel

1Introduction

Searchandrescueindisasterareainthedomainofrobotisaburgeoningandchallengingsubject[1].Minerescuerobotwasdevelopedtoenterminesduringemergenciestolocatepossibleescaperoutesforthosetrappedinsideanddeterminewhetheritissafeforhumantoenterornot.Localizationisafundamentalprobleminthisfield.Localizationmethodsbasedoncameracanbemainlyclassifiedintogeometric,topologicalorhybridones[2].Withitsfeasibilityandeffectiveness,scenerecognitionbecomesoneoftheimportanttechnologiesoftopologicallocalization.

Currentlymostscenerecognitionmethodsarebasedonglobalimagefeaturesandhavetwodistinctstages:

trainingofflineandmatchingonline.

Duringthetrainingstage,robotcollectstheimagesoftheenvironmentwhereitworksandprocessestheimagestoextractglobalfeaturesthatrepresentthescene.Someapproacheswereusedtoanalyzethedata-setofimagedirectlyandsomeprimaryfeatureswerefound,suchasthePCAmethod[3].However,thePCAmethodisnoteffectiveindistinguishingtheclassesoffeatures.Anothertypeofapproachusesappearancefeaturesincludingcolor,textureandedgedensitytorepresenttheimage.Forexample,ZHOUetal[4]usedmultidimensionalhistogramstodescribeglobalappearancefeatures.Thismethodissimplebutsensitivetoscaleandilluminationchanges.Infact,allkindsofglobalimagefeaturesaresufferedfromthechangeofenvironment.

LOWE[5]presentedaSIFTmethodthatusessimilarityinvariantdescriptorsformedbycharacteristicscaleandorientationatinterestpointstoobtainthefeatures.Thefeaturesareinvarianttoimagescaling,translation,rotationandpartiallyinvarianttoilluminationchanges.ButSIFTmaygenerate1000ormoreinterestpoints,whichmayslowdowntheprocessordramatically.

Duringthematchingstage,nearestneighborstrategy（NN）iswidelyadoptedforitsfacilityandintelligibility[6].Butitcannotcapturethecontributionofindividualfeatureforscenerecognition.Inexperiments,theNNisnotgoodenoughtoexpressthesimilaritybetweentwopatterns.Furthermore,theselectedfeaturescannotrepresentthescenethoroughlyaccordingtothestate-of-artpatternrecognition,whichmakesrecognitionnotreliable[7].

Sointhisworkanewrecognitionsystemispresented,whichismorereliableandeffectiveifitisusedinacomplexmineenvironment.Inthissystem,weimprovetheinvarianeebyextractingsalientlocalimageregionsasIandmarkstoreplacethewholeimagetodealwithlargechangesinscale,2Drotationandviewpoint.Andthenumberofinterestpointsisreducedeffectively,whichmakestheprocessingeasier.FuzzyrecognitionstrategyisdesignedtorecognizetheIandmarksinplaceofNN,whichcanstrengthenthecontributionofindividualfeatureforscenerecognition.Becauseofitspartialinformationresumingability,hiddenMarkovmodelisadoptedtoorganizethoselandmarks,whichcancapturethestructureorrelationshipamongthem.SoscenerecognitioncanbetransformedtotheevaluationproblemofHMM,whichmakesrecognitionrobust.

2Salientlocalimageregionsdetection

Researches）nbiologicalvisionsystemindicatethatorganism（likedrosophila）oftenpaysattentiontocertainspecialregionsinthescenefortheirbehavioralrelevanceorlocalimagecueswhileobservingsurroundings[8].TheseregionscanbetakenasnaturalIandmarkstoeffectivelyrepresentanddistinguishdifferentenvironments.Inspiredbythose,weusecenter-surrounddiffereneemethodtodetectsalientregionsinmulti-scaleimagespaces.Theopponenciesofcolorandtexturearecomputedtocreatethesaliencymap.

Follow-up,sub-imagecenteredatthesalientpositioninSistakenastheIandmarkregion.ThesizeoftheIandmarkregioncanbedecidedadaptivelyaccordingtothechangesofgradientorientationofthelocalimage[11].

Mobilerobotnavigationrequiresthatnaturallandmarksshouldbedetectedstablywhenenvironmentschangetosomeextent.TovalidatetherepeatabilityonIandmarkdetectionofourapproach,wehavedonesomeexperimentsonthecasesofscale,2Drotationandviewpointchangesetc.Fig.1showsthatthedoorisdetectedforitssaliencywhenviewpointchanges.Moredetailedanalysisandresultsaboutscaleandrotationcanbefoundinourpreviousworks[12].

3Scenerecognitionandlocalization

Differentfromotherscenerecognitionsystems,oursystemdoesn'needtrainingoffline.Inotherwords,ourscenesarenotclassifiedinadvance.Whenrobotwanders,scenescapturedatintervalsoffixedtimeareusedtobuildthevertexofatopologicalmap,whichrepresentstheplacewhererobotlocates.Althoughthemapgeometriclayoutisignoredbythelocalizationsystem,itisusefulforvisualizationanddebugging[13]andbeneficialtopathplanning.Solocalizationmeanssearchingthebestmatchofcurrentseeneonthemap.InthispaperhiddenMarkovmodelisusedtoorganizetheextractedIandmarksfromcurrentsceneandcreatethevertexoftopologicalmapforitspartialinformationresumingability.

Resembledbypanoramicvisionsystem,robotlooksaroundtogetomni-images.From

Fig.1Experimentonviewpointchanges

eachimage,salientlocalregionsaredetectedandformedtobeasequenee,namedasIandmarksequencewhoseorderisthesameastheimagesequence.ThenahiddenMarkovmodeliscreatedbasedontheIandmarksequenceinvolvingksalient

localimageregions,whichistakenasthedescriptionoftheplacewheretherobotlocates.InoursystemEVI-D70camerahasaviewfieldof±70°Consideringtheoverlapeffect,wesampleenvironmentevery45°toget8images.

Letthe8imagesashiddenstateSi（1

Asfortheedgeoftopologicalmap,weassignitwithdistanceinformation

betweentwovertices.Thedistancescanbecomputedaccordingtoodometryreadings.

Fig.2HMMofenvironment

Tolocateitselfonthetopologicalmap,robotmustrunits‘eyeonenvironment

andextractaIandmarksequenceL1'-Lk',thensearchthemapforthebestmatchedvertex（scene）.Differentfromtraditionalprobabilisticlocalization[15],inoursystemlocalizationproblemcanbeconvertedtotheevaluationproblemofHMM.Thevertexwiththegreatestevaluationvalue,whichmustalsobegreaterthanathreshold,istakenasthebestmatchedvertex,whichindicatesthemostpossibleplacewheretherobotis.

4Matchstrategybasedonfuzzylogic

Oneofthekeyissuesinimagematchproblemistochoosethemosteffectivefeaturesordescriptorstorepresenttheoriginalimage.Duetorobotmovement,thoseextractedlandmarkregionswillchangeatpixellevel.So,thedescriptorsorfeatureschosenshouldbeinvarianttosomeextentaccordingtothechangesofscale,rotationandviewpointetc.Inthispaper,weuse4featurescommonlyadoptedinthecommunitythatarebrieflydescribedasfollows.

GO:

Gradientorientation.Ithasbeenprovedthatilluminationandrotationchangesarelikelytohavelessinfluenceonit[5].

ASMandENT:

Angularsecondmomentandentropy,whicharetwotexturedescriptors.

H:

Hue,whichisusedtodescribethefundamentalinformationoftheimage.

Anotherkeyissueinmatchproblemistochooseagoodmatchstrategyoralgorithm.Usuallynearestneighborstrategy（NN）isusedtomeasurethesimilaritybetweentwopatterns.ButwehavefoundintheexperimentsthatNNcan'atequatelyexhibittheindividualdescriptororfeaturecontributiontosimilaritymeasurement.AsindicatedinFig.4,theinputimageFig.4（a）comesfromdifferentviewofFig.4（b）.ButthedistancebetweenFigs.4（a）and（b）computedbyJeffereydivergenceislargerthanFig.4（c）.

Tosolvetheproblem,wedesignanewmatchalgorithmbasedonfuzzylogicforexhibitingthesubtlechangesofeachfeatures.Thealgorithmisdescribedasbelow.

AndtheIandmarkinthedatabasewhosefusedsimilaritydegreeishigherthananyothersistakenasthebestmatch.ThematchresultsofFigs.2（b）and（c）aredemonstratedbyFig.3.Asindicated,thismethodcanmeasurethesimilarity

effectivelybetweentwopatterns.

o48121620

Landmarkindatabase（index）

Fig.3Similaritycomputedusingfuzzystrategy

5Experimentsandanalysis

Thelocalizationsystemhasbeenimplementedonamobilerobot,whichisbuiltbyourlaboratory.ThevisionsystemiscomposedofaCCDcameraandaframegrabberIVC-4200.Theresolutionofimageissettobe400>320andthesamplefrequencyissettobe10frames/s.Th