文献翻译——在苹果树冠的超声波测距传感器的性能

0外文文献资料PerformanceofanUltrasonicRangingSensorinAppleTreeCanopiesAbstract:Electroniccanopycharacterizationisanimportantissueintreecropmanagement.Ultrasonicandopticalsensorsarethemostusedforthispurpose.Theobjectiveofthisworkwastoassesstheperformanceofanultrasonicsensorunderlaboratoryandfieldconditionsinordertoprovidereliableestimationsofdistancemeasurementstoappletreecanopies.Tothispurpose,amethodologyhasbeendesignedtoanalyzesensorperformanceinrelationtofoliagerangingandtointerferenceswithadjacentsensorswhenworkingsimultaneously.Resultsshowthattheaverageerrorindistancemeasurementusingtheultrasonicsensorinlaboratoryconditionsis0.53cm.However,theincreaseofvariabilityinfieldconditionsreducestheaccuracyofthiskindofsensorswhenestimatingdistancestocanopies.Theaverageerrorinsuchsituationsis5.11cm.Whenanalyzinginterferencesofadjacentsensors30cmapart,theaverageerroris17.46cm.Whensensorsareseparated60cm,theaverageerroris9.29cm.Theultrasonicsensortestedhasbeenproventobesuitabletoestimatedistancestothecanopyinfieldconditionswhensensorsare60cmapartormoreandcould,therefore,beusedinasystemtoestimatestructuralcanopyparametersinprecisionhorticulture.Keywords:ultrasonicsensor;distancemeasurements;appletreeorchard;ultrasonicinterferences1.IntroductionUltrasonicsensorshavebeenusedformanypurposesinagricultureformorethan40years.Oneoftheseapplicationshasbeenthedetectionandrangingtoextractgeometricinformationfromfruittreecanopies.Thefirstdevelopmentsinthisareawererelatedtotheapplicationofplantprotectionproductssuchaspesticidesandfungicidesinfruitorchards.Oncedosesstartedtobeadjustedaccordingtotheamountofvegetationtobetreated1,2,someresearchersbegantodevelopelectronicsystemstoquantifythegeometricparametersofcanopies.Thefirstproposalstoestimatecanopyvolumeusedseveralultrasonicsensorsmountedinaverticalmast3ormountedonasprayer4drivenbyatractor.However,thestate-of-the-artoftheapplicationtechnologiesdidnotallowthisinformationtobeusedinrealtime.Instead,somesprayerpatents5,6andscientificworks7-10werepublishedmountingultrasonicsensorstoonlydetectthepresenceofcanopyinordertoexclusivelyspraywhenvegetationwasinfrontofthenozzles.Anotherapplicationofultrasonicsensorsistheonedesignedtospraycitruscanopiesataconstantgiven1distance11.Thenozzlesaremountedonamovingarmcontrolledtofollowthecontourofthecanopyaccordingtotheinformationprovidedbyasensor.Inthisworkanaccuracyanalysiscanbefoundwhensensorsareplaced50cmand75cmapart.Theaverageerrorintheircitrusgrovedistancemeasurementsis11.40cm.However,informationisnotprovidedabouttheimportanceoftheeffectofthefoliageontheultrasonicconeortheeffectofinterferencesinsuchanerror.Thesameauthorsdevelopedasprayerabletospraythreedifferentflowratesaccordingtoacanopywidthestimationmadewithanultrasonicsensor12.Therewasnosprayingwhentherewasnovegetation;halftheflowratewassprayedwhenlittlevegetationwasdetectedandthefullflowratewassprayedwhencanopywidthwashigherthanapredefinedthreshold.Thisdesignledthewaytoacontinuousvariationoftheflowrateaccordingtothevariabilityofthevegetationalongfruitorchard,vineyardorcitrusgroverows13-18.Alotofresearchhasbeencarriedouttoautomaticallymeasurecanopydimensionsincitrus.Firstworkswerefocusedoncomparingmanualvolumeestimationswithmeasurementsperformedwithultrasonicandlidarsensors19.Resultsshowedthattheestimationsofultrasonicandlidarsensorscorrelatedfairlywell,whereasthecorrelationwithmanualestimationswaslower.Theauthorsattributedthedifferenceswithmanualmeasurementstothehighersamplingresolutionachievedwiththesensors.Theultrasonicsystemconsistedofamastwith10ultrasonicsensorsperside.Inordertoavoidsignalinterferences,alternatedsensorswerefiredindifferentgroupssequentially.ThissystemwaslaterusedtomapcanopyvolumesincitrusorchardsbyfittingaDGPSreceiver20-22.Theauthorsobservedthatinlessdensetreesbiggerdifferenceswerefoundbetweenmanualandsensorestimations.Theseauthorsusedcanopyvolumeinformationtoadjustthefertilizerdoserate23,24andtoestimatethefruityield.Inrelationtotheverticalsamplingresolution,lidarsensorsprovidemuchmoreinformationthananarrayofultrasonicsensorsresultinginamoreaccurateestimationofcanopyparameters26-29.Anotherapplicationhasbeentheuseofultrasonicallyestimatedcanopyvolumes,togetherwithotherinformation,todefinemanagementzonesincitrusgroves30,31.Sourcesoferrorsinestimatingcanopyvolumesincitruswiththepreviouslydescribedsystemincomparisonwithmanualmeasurementshavealsobeenstudied32.ThemostimportantsourceoferrorintheirsystemwastheinaccuracyoftheDGPSreceiverinestimatinggroundspeedfollowedbytheeffectofairtemperatureonthetimeofflightofsoundwavesand,therefore,onthesensorsdistanceestimation.Alastsourceoferrorwasthedeviationsinthedrivingpath.Nonetheless,noinformationwasprovidedabouttheeffectofcanopystructureonthesensoroutputorotherparameterspotentiallyaffectingthereflectanceoftheultrasonicwaves.Recently,anultrasonicsensorhasbeenstudiedinordertodetermineitsaccuracywhenmeasuringdistancestosimulatedcanopiesoffieldcrops33.Theresultsarepromisingforfieldcropsbutdifficulttoextrapolatetotreecropsduetodifferencesincanopystructurebetweencrops.Mostofthereferencedapplicationsofultrasonicsensorsincanopycharacterizationarefocusedoncorrelatingmanualestimationsofwidthorvolumewiththeresultsobtainedbyusingthesensors.Theseworksdidnotprovideinformationabouttheinteractionbetweenthesoundwaveandthecanopyitselfandhowcanitinterfereintheestimationsofultrasonicsensors.Inrangingapplications,ultrasonicsensorsareintendedtoestimatedistancestoobjectswithsolidsurfacesinordertogetspecularreflectionsoftheacousticwaves.Thesurfaceofanappletreecanopyismadeupofsmallleafsurfacesplacedindifferentorientations.Consequently,thereflectionofultrasonicwavesismore2diffusethanspecularandthiscanstronglyaffectestimateddistances.Inthispaperacommercialultrasonicsensormodelisanalyzedtovalidateitssuitability,performanceandreliabilityinanappleorchardandforabetterunderstandingofthesensingprocess.Trialshavebeenperformedinlaboratoryandfieldconditionsinastationarywayinordertoassessitsabilitytoestimatebothdistancestothecanopyandtheeffectofpossibleinterferencescomingfromotheradjacentsensors.Thisstudyisthefirststepinalargerworkthatwouldusethistypeofsensorstoestimatecanopyparameterssuchascrosssectionareasorcanopyvolumeinfruittreecropsintheframeworkofprecisionhorticulture.2.MaterialsandMethodsOneultrasonicsensorandadataacquisitionsystemwereusedinthelaboratoryandfieldtrials.Thesensorsweremountedonaverticalaluminummastonamobileplatform,whichwasalsousedtocarrytheacquisitionsystemandthebatteriestosupplytherequiredvoltagetoruntheelectronicdevices.2.1.SensorandDataAcquisitionTheultrasonicsensorselectedtocarryoutthetrialsisaSonarBeroPXS400M30K3(SiemensAG,Munich,Germany,Figure1left).Adifferentmodelofthesameserieshasalreadybeenusedinfieldworksincitruswithanincreaseddetectionlimit11.Thesensorisconsideredtoberobustenoughintermsofworkingtemperature(thesensorhasaninternaltemperaturesensortocompensatetheeffectofthisparameterinthedistanceestimation),degreeofprotection,shockandvibrationresistance.Sincerowspacinginfruittreeorchardsisusually5morlessandthesensoristhoughttobeplacedinthecenterofthealleys,thechosensensingrangewasfrom40cmto300cm.However,anysensingrangemodificationresultsindifferencesintheemittedultrasonicconeangle:theshortertherange,thenarrowertheultrasonicconeangle.Duetothisdifferencewiththesensorusedincitrusandthedifferencesinthecropitself,itwasconsideredimportanttoassesstheperformanceofthissensorinanappleorchardbeforebeingusedincanopycharacterization.OtherspecificationsofthesensorcanbefoundinTable1.Theacquisitionsystem(Figure1right)consistsofaPAC(ProgrammableAutomationController)modelcompactFieldPoint(NationalInstruments,Austin,TX,USA).Thisdevicemountsananaloginputmodulewherethesensorsareconnected.ThisdevicecanbeoperatedfromalaptoprunningspecificallydesignedpiecesofsoftwareprogrammedusingLabVIEW(NationalInstruments).Animportantparametertobetakenintoaccountisthebeamangle.Inthespecificationsheetdeliveredtogetherwiththesensor,theinformationprovidedrelatedtothisparameterisasfollows:intheoperatingrange,objectsaresensedreliablywithinasoundconeangleof5.Undergoodreflectionconditions,objectscanalsobesensedoutside.Inamoredetailed3catalog,themanufacturerprovidesinformationabouttheshapeandsizeofthesoundconesdependingontheobjecttobedetected(Figure2).Thetargetsarea1010cmsquaredplaneobjectandan8cmdiametercylindricalobject.Toobtainthesediagrams,thesquaredtargetisplacedbothaligned,withthemostoptimumreflection(Figure2(a),andperpendicularlytotheultrasonicconeaxis(Figure2(c).ThecylindricaltargetconediagramisdescribedinFigure2(b).Anultrasonicconeprojectionwitha5beamanglehasbeensuperimposedtothediagrams.Itisclearthatthetargetitselfanditsrelativepositiontothesensorcangreatlyaffectthedetectionfootprintofthesensor.Ultrasonicsensorsarealsoverysensitivetointerferingsonicwavescomingfromnearbysensors.Themanufacturerofferstwomethodstosynchronizeseveralsensorsinordertoavoidinaccuratereadingsduetointerferencesofsonicechoessentfromanothersensor.However,bothmethodspresentbigdisadvantages.Inonemethodtheobjectcannotbeassignedtoaparticularsensor.Intheother,longerresponsetimesarerequiredbecauseeachsensorisonlyactivebrieflyandthenhastowaituntilalltheothersensorsinthecircuithaveemitted.Thelatter,causesthearrayofsensorsnottoobtain.simultaneousreadings,whatcouldcauseinaccurateestimationsofcross-sectionalareasoftreerows.Thesesolutionscouldbeusefulinastationarysystembutarenotsatisfyingatallwhenthesensorsareboardedinamovingplatform.Inthislastsituation,aminimumdistancebetweenactivesensorsmustbefoundinordertoobtainthehighestverticalresolutiontobetterestimatecanopyparameterswiththeleasteffectofinterferences.2.2.LaboratoryDistanceMeasurementTrialThelaboratorydistancemeasurementtrialwascarriedoutattheCentredeMecanitzaciAgrria(CenterofAgriculturalMechanization)oftheDepartmentofAgriculture,Livestock,Fisheries,FoodandEnvironmentoftheGeneralitatdeCatalunya(CatalanGovernment)inLleida,Catalonia,Spain.Thetrialwasdesignedtoassessthecapabilitiesoftheultrasonicsensorunderidealconditionsusingahard,flat,10cm100cmmetallictarget(biggerthantheminimumdimensionsrecommendedbythemanufacturer).Themobileplatformcarryingthesensorwasstationedandthetargetwasplacedatdifferentknowndistancestothesensor(Figure3).2,000readingswerestoredateachspecificdistancerangingfrom45cmto285cminstepsof15cm.Allmeasurementswereperformedinstationaryandnowindconditions.Thestatisticalanalysisconsistedoffittingalinearregressionmodel(Equation(1)inordertoestablishtheexistenceofafunctionalrelationshipbetweenthedistancetothesensorandthesensoranalogoutput.Ifpossible,aregressionlinewouldbedefinedaccordingtotheleastsquaresmethod.Thequalityofthefittingwasassessedbyanalyzingthemeasureddistancevs.thesensoroutputandthemeasureddistancevs.theresidualscatterdiagramsandbyanalyzingthecoefficientsof4correlation(r)anddetermination(R2),therootmeansquareerror(RMSE)andthesignificanceofthemodel.3.ResultsandDiscussion3.1.LaboratoryDistanceMeasurementTrialThetotalnumberofobservationswas34,000.However,fifteenofthemwereremovedastheywereconsideredoutliers.InTable3itisclearlyseenthatthecorrelationbetweentheoutputsignalofthesensorandthetargetdistanceisverystrong.Thefittedmodelcanexplaina99.9%ofthevariabilityoftheresponse.TheRMSEisverylowandgivesanideaofthegoodfittingofthemodel.Thesignificanceofthefittedmodel,representedbythepvalueoftheanalysisofvarianceofthemodel,isveryhigh.Thepvalueleadstorefusethenullhypothesiswhichassumesthatallparametersareequaltozero.Thereforeitispossibletofindasignificantparameterfortheregressionline(Equation(2).d=297.0726.36v(2)wheredistheestimateddistance,expressedincmandvistheoutputvoltagesignalofthesensor,expressedinV.3.2.FieldTrials3.2.1.DistanceMeasurementTrialTotalnumberofobservationswas168.Tenobservationswereremovedastheywereconsideredabnormalwhencomparedwiththeirequivalentmeasurementtothe10cm2cardboardtarget.InTable4itisseenthatthecorrelationbetweenbothvariablesisstillverystrong.Thefittedmodelcanexplain98%ofthevariabilityofthedistancetothefirstleaf.ThemaindifferencewiththemodelfittedunderlaboratoryconditionsistheRMSE.Whileanartificialplanetargetproducesagoodecho,theincreaseofRMSEinfieldconditionstellsthatvariabilityintheresponseishigher.Eventhough,thefittedmodelisstillhighlysignificant,asstatedbythepvalue.Soitispossibletofindasignificantparameterfortheregressionline(Equation(3).WhencomparingEquations(2)and(3),thedifferencebetweenthetwoparametersisrathersmallwhileinterceptsdiffer2.79cm.Theregressionline,aswellasresultsofthedistancemeasurementlaboratorytrial,isshowninFigure8.Theresidualscatterplotdemonstratesthatthereisnotaclearstructurefortheresidualssowecanacceptthefittedmodel.5d=294.2825.82v(3)wheredistheestimateddistance,expressedincmandvistheoutputvoltagesignalofthesensor,expressedinV.3.2.2.InterferenceTrialAtotalof113observationswerecarriedoutaccordingtothemethodologydescribedinthefluxdiagraminFigure6(b).Adjacentsensorsplacedat30cmand60cmdointerferetheoutputsignalofthecentralsensor,asseeninFigure11.Whenusingsensorsseparated60cm,thefirst100cmarelesssensitivetointerferencesthantherestofthesensingrange,whiletheinterferencescausedbysensors.At30cmaffectallthesensingrange.Ingeneral,thefurtherthecanopy,thebiggertheeffectsprovokedbynearbysensorsare.Theconsequenceofinterferencesisthattheoutputsignalofinterferedsensorsmaysporadicallybehigherthanitshouldbe.Accordingtothefielddistanceestimation(Equation(3)interferenceswouldtendtodecreasethepredicteddistancetothecanopywhichimpliesanoverestimationofitswidth.Inthecaseofusingultrasonicsensorstodeterminethecanopyvolumetoadjustthedoserateofplantprotectionproducts,overestimationofcanopyvolumeswithinterferencesisnotasnegativeasunderestimation.Underestimationscoulddecreasetheefficacyofthetreatmentandcreateresistanceofpesttocertainactiveingredients.4.ConclusionsThetestedultrasonicsensorisabletoaccuratelyestimatedistancesunderlaboratoryconditionswithanaverageerrorof0.53cm.Whenusedunderfieldconditions,thedistanceestimationequationshouldbeadaptedtobetterestimatedistancestothecanopy.However,differenceswiththelaboratoryestimationequationarerelativelysmall,consideringotherpossiblesourcesoferror.Thevariabilityindistanceestimationsinfieldconditionsinanappleorchardclearlyincreasesinrelationtowhatwasobtainedinlaboratorywithartificialtargets.Asaconsequenceofthis,theaverageerroris5.11cm.Theeffectofinterferencesishigherwhensensorsare30cmapartwithanaverageerrorof17.46cm.Whensensorsareseparated60cm,theaverageerroris9.29cm.Sensorsshouldthusbeseparatedmorethan60cminordertoavoidhighinterferenceeffects.Ultrasonicsensorsliketheonetestedandreportedinthispaperhavebeenproventobesuitabletoestimatedistancestothecanopyinfieldconditions.Resultscouldbeextrapolatedtootherapplecropvarietiesandotherspeciessuchaspearcropswherecanopystructuresandleafdimensionsaresimilar.However,ithastobetakenintoaccountthattheincreaseofvariabilityduetothecharacteristics6ofthecanopysurfaceandtheultrasonicworkingprinciplereducestheaccuracyoftheestimationsandthattheeffectofinterferencescanbeimportantwhenadjacentsensorsaretooclose.References1.Morgan,N.G.Gallonsperacreofsprayedareaanalternativestandardtermforthesprayingofplantationcrops.WorldCrop1964,16,64-65.2.Byers,R.E.;Hickey,K.D.;Hill,C.H.Basegallonageperacre.VirginiaFruit1971,60,19-23.3.McConnell,R.L.;Elliot,K.C.;Blizzard,S.H.;Koster,K.H.Electronicmeasurementoftreerowvolume.Agr.Electron.1983,1,85-90.4.Giles,D.K.;Delwiche,M.J.;Dodd,R.B.Electronicmeasurementoftreecanopyvolume.Trans.ASABE1988,31,264-272.5.Roper,B.E.GroveSprayer.U.S.Patent4,768,713,6September1988.6.Giles,D.K.;Delwiche,M.J.;Dodd,R.B.MethodandApparatusforTargetPlantFoliageSensingandMappingandRelatedMaterialsApplicationControl.U.S.Patent4,823,268,18April1989.7.Giles,D.K.;Delwiche,M.J.;Dodd,R.B.Controloforchardsprayingbasedonelectronicsensingoftargetcharacteristics.Trans.ASABE1987,30,1624-1630,1636.8.Giles,D.K.;Delwiche,M.J.;Dodd,R.B.Sprayercontrolbysensingorchardcropcharacteristics:Orchardarchitectureandsprayliquidsavings.J.Agr.Eng.Res.1989,43,271-289.9.Balsari,P.;Tamagnone,M.Anautomaticspraycontrolforairblastsprayers:Firstresults.InPreciscionAgriculture97;InProceedingsofthe1stEuropeanConferenceonPrecisionAgriculture,Warwick,UK,September1997;Stafford,J.V.,Ed.;BIOSScientificPublishersLtd.:Oxford,UK,1997;pp.619-626.10.Brown,D.L.;Giles,D.K.;Oliver,M.N.;Klassen,P.Targetedspraytechnologytoreducepesticideinrunofffromdormantorchards.CropProt.2008,27,545-552.11.Molt,E.;Martn,B.;Gutirrez,A.Designandtestingofanautomaticmachineforsprayingataconstantdistancefromthetreecanopy.J.Agr.Eng.Res.2000,77,379-384.7中文翻译稿在苹果树冠的超声波测距传感器的性能摘要：电子树冠特征是林木管理的一个重要课题。超声和光学传感器是最常用的用于此目的。这项工作的目的是评估，以提供距离测量的可靠估计到苹果树檐实验室和现场条件下的超声波传感器的性能。为此目的，一种方法已被设计来分析相对于传感器性能叶面同时工作时测距和对干扰与相邻的传感器。结果表明，使用了超声波传感器在实验室条件下在距离测量的平均误差是0.53公分。然而，变异性的现场条件的增加估计的距离来篷时减少这种传感器的精度。在这种情况下的平均误差为5.11厘米。在分析相邻传感器的干扰30厘米分开，平均误差为17.46厘米。当传感器被隔开60厘米，平均误差为9.29厘米。超声波传感器检测已被证明是合适的估计的距离，以在现场条件篷当传感器60厘米外或多个和可能，因此，可以使用在一个系统来估计园艺结构参数精度。关键词：超声波传感器;距离测量;苹果树果园;超声波干扰1.简介超声波传感器已被用于许多目的的农业超过40年。这些应用一直是用在探测和测距提取果树檐几何信息。在这一领域的第一个发展均与植物保护产品如杀虫剂和杀真菌剂在果园中的应用有关。一旦剂量开始，根据植物的待处理量1,2进行调整，一些研究人员开始发展的电子系统来量化檐的几何参数。第一方案来估计树冠量使用安装在垂直桅杆3或安装在喷雾器几个超声波传感器4驱动拖拉机。然而，先进设备，先进的应用技术并没有让实时使用这些信息。相反，一些喷雾器专利5,6和科学著作7-10发表安装超声波传感器只检测以便喷洒时植物是在喷嘴的前方篷存在。超声波传感器的另一个应用是一个设计成喷柑桔檐以恒定给定的距离11。喷嘴被安装在一移动臂控制根据由传感器提供的信息来遵循篷的轮廓。在这项工作中的精度分析可以发现，传感器需被放置50厘米和75厘米外。在他们的柑橘林中距离测量平均误差为11.40厘米。然而，在没有提供关于枝叶上的超声波锥体或干扰的情况下，这样的错误是影响效果的重要信息。作者同时开发能够根据与超声波传感器12制成篷宽度估计以三个不同流率的喷雾器。喷洒在没有植被;一半的流速喷洒，检测小植被时和充分流速喷雾时篷宽度比预定阈值高。根据果园，葡8萄园或柑桔林行13-18的植被的变异，这种设计导致的方式上流量的连续变化。大量的研究已经进行了自动测量柑橘树冠尺寸。第一个作品都集中在比较手动量估计与超声波和激光雷达传感器19进行的测量。结果表明，超声波和激光雷达传感器估计相关还算不错，而采用手动估计的相关性较低。笔者归结差异与手工测量有关，以实现与传感器更高的采样分辨率。超声波系统包括桅杆的每面10的超声波传感器。为了避免信号干扰，传感器交替依次发射了不同的组。这个系统后来被拟合DGPS接收机20-22映射在柑橘园树冠体积。笔者观察到，在密度较小的树木手动和传感器之间的估算发现更大的差异。这些差异作者使用树冠量信息来调整肥料剂量率23,24，并以估计的果实产量。相对于垂直采样分辨率，激光雷达传感器提供比导致冠层参数26-29的更准确的估计的超声波传感器阵列更多的信息。另一个应用程序已使用超声估计树冠体积，与其他信息一起，在柑橘园30,31定义的管理区。在估计篷卷柑桔与先前描述的系统与手动测量比较错误的来源也进行了研究32。在他们的系统误差的最重要来源是DGPS接收机估计地面速度，随后空气温度对声波的速度，因此，在传感器距离估计的时间的影响是不