机器学习算法在道路质量监测卫星图像上的应用：道路质量调查的另一种方法（英）

APPLICATIONOFMACHINELEARNINGALGORITHMSONSATELLITEIMAGERYFOR ROADQUALITYMONITORING ANALTERNATIVEAPPROACHTOROADQUALITYSURVEYS AaronThegeya,ThomasMitterling,ArturoMartinezJr,JosephAlbertNiñoBulan,RonLesterDurante,andJayzonMag-atas NO.675 December2022 ADBECONOMICSWORKINGPAPERSERIES ASIANDEVELOPMENTBANK ADBEconomicsWorkingPaperSeries ApplicationofMachineLearningAlgorithms onSatelliteImageryforRoadQualityMonitoring:AnAlternativeApproachtoRoadQualitySurveys AaronThegeya,ThomasMitterling, ArturoMartinezJr,JosephAlbertNiñoBulan,RonLesterDurante,andJayzonMag-atas TheADBEconomicsWorkingPaperSeriespresentsresearchinprogresstoelicitcommentsandencouragedebateondevelopmentissuesinAsiaandthePacific.TheviewsexpressedarethoseoftheauthorsanddonotnecessarilyreflecttheviewsandpoliciesofADBoritsBoardofGovernorsorthegovernmentstheyrepresent. No.675|December2022 AaronThegeya(athegeya@gmail.com)andThomasMitterling(thomas.mitterling@worlddata.io)aredatascientistsattheWorldDataLab.ArturoMartinezJr.isastatistician(amartinezjr@adb.org),JosephBulan (jbulan@adb.org)isanassociatestatisticsanalyst,andRonLesterDurante(rdurante1.consultant@adb.org)andJayzonMag-atas(jmagatas.consultant@adb.org)areconsultantsattheEconomicResearchandRegionalCooperationDepartment,AsianDevelopmentBank. ASIANDEVELOPMENTBANK CreativeCommonsAttribution3.0IGOlicense(CCBY3.0IGO) ©2022AsianDevelopmentBank 6ADBAvenue,MandaluyongCity,1550MetroManila,PhilippinesTel+63286324444;Fax+63286362444 www.adb.org Somerightsreserved.Publishedin2022. ISSN2313-6537(print),2313-6545(electronic)PublicationStockNo.WPS220587-2 DOI:http://dx.doi.org/10.22617/WPS220587-2 TheviewsexpressedinthispublicationarethoseoftheauthorsanddonotnecessarilyreflecttheviewsandpoliciesoftheAsianDevelopmentBank(ADB)oritsBoardofGovernorsorthegovernmentstheyrepresent. ADBdoesnotguaranteetheaccuracyofthedataincludedinthispublicationandacceptsnoresponsibilityforanyconsequenceoftheiruse.ThementionofspecificcompaniesorproductsofmanufacturersdoesnotimplythattheyareendorsedorrecommendedbyADBinpreferencetoothersofasimilarnaturethatarenotmentioned. Bymakinganydesignationoforreferencetoaparticularterritoryorgeographicarea,orbyusingtheterm“country”inthisdocument,ADBdoesnotintendtomakeanyjudgmentsastothelegalorotherstatusofanyterritoryorarea. ThisworkisavailableundertheCreativeCommonsAttribution3.0IGOlicense(CCBY3.0IGO)https://creativecommons.org/licenses/by/3.0/igo/.Byusingthecontentofthispublication,youagreetobeboundbythetermsofthislicense.Forattribution,translations,adaptations,andpermissions,pleasereadtheprovisionsandtermsofuseathttps://www.adb.org/terms-use#openaccess. ThisCClicensedoesnotapplytonon-ADBcopyrightmaterialsinthispublication.Ifthematerialisattributedtoanothersource,pleasecontactthecopyrightownerorpublisherofthatsourceforpermissiontoreproduceit.ADBcannotbeheldliableforanyclaimsthatariseasaresultofyouruseofthematerial. Pleasecontactpubsmarketing@adb.orgifyouhavequestionsorcommentswithrespecttocontent,orifyouwishtoobtaincopyrightpermissionforyourintendedusethatdoesnotfallwithintheseterms,orforpermissiontousetheADBlogo. CorrigendatoADBpublicationsmaybefoundathttp://www.adb.org/publications/corrigenda.Note: Inthispublication,“$”referstoUnitedStatesdollars. TheADBEconomicsWorkingPaperSeriespresentsdata,information,and/orfindingsfromongoingresearchandstudiestoencourageexchangeofideasandtoelicitcommentandfeedbackaboutdevelopmentissuesinAsiaandthePacific.Sincepapersinthisseriesareintendedforquickandeasydissemination,thecontentmayormaynotbefullyeditedandmaylaterbemodifiedforfinalpublication. ABSTRACT Roadsarevitaltosupportthetransportationofpeople,goods,andservices,amongothers.Toyieldtheiroptimalsocioeconomicimpact,propermaintenanceofexistingroadsisrequired;however,thisistypicallyunderfunded.Sincedetectingroadqualityisbothlaborandcapitalintensive,informationonitisusuallyscarce,especiallyinresource-constrainedcountries.Accordingly,thestudyexaminesthefeasibilityofusingsatelliteimageryandartificialintelligencetodevelopanefficientandcost-effectivewaytodetermineandpredicttheconditionofroads.Withthisgoal,apreliminaryalgorithmwascreatedandvalidatedusingmedium-resolutionsatelliteimageryandexistingroadroughnessdatafromthePhilippines.Afteranalysis,itwasdeterminedthatthealgorithmhadanaccuracyrateupto75%andcanbeusedforthepreliminaryidentificationofpoortobadroads.Thisprovidesanalternativeforcompilingroadqualitydata,especiallyforareaswhereconventionalmethodscanbedifficulttoimplement.Nonetheless,additionaltechnicalenhancementsneedtobeexploredtofurtherincreasethealgorithm’spredictionaccuracyandenhanceitsrobustness. Keywords:roadquality,roadmaintenance,SustainableDevelopmentGoals,remotesensing,deeplearning JELcodes:O18,R42 ThispaperwaspreparedforthetechnicalassistanceprojectfundedbytheJapanFundforProsperousandResilientAsiaandthePacificentitledUsingFrontierTechnologyandBigDataAnalyticsforSmartInfrastructureFacilityPlanningandMonitoring.TheauthorswouldliketothankcolleaguesfromthePhilippineDepartmentofPublicWorksandHighw