Short
communication
Validation
of
an
indirect
data
collection
method
to
assess
airport
pavement
condition
Bertha
Santos
a,*
,
Pedro
G.
Almeida
b,
Ianca
Feitosa
c,
Débora
Lima
daUniversityofBeiraInterior,DepartmentofCivilEngineeringandArchitecture,6200-358,Covilhã,PortugalandCERIS(ECI/04625),Civil
EngineeringResearchandInnovationforSustainability,Lisbon,Portugal
b
UniversityofBeiraInterior,DepartmentofCivilEngineeringandArchitectureandGEOBIOTEC(GEO/04035),GeoBioSciences, GeoTechnologiesandGeoEngineering,6200-358,Covilhã,Portugal
c
UniversityofBeiraInterior,DepartmentofCivilEngineeringandArchitecture,6200-358,Covilhã,Portugal
d
CapeVerdeAirportsandAirSafety(ASA,SA),SalIsland,CapeVerdeandUniversityofBeiraInterior,DepartmentofCivilEngineeringand Architecture,6200-358,Covilhã,Portugal
ARTICLE INFO Articlehistory: Received11March2020
Receivedinrevisedform28July2020 Accepted5August2020
Keywords:
Airportpavementmanagementsystem Datacollection
Imageprocessing Statisticalcomparison
ABSTRACT
Inthisstudytheauthorscomparetwomethodsforairportasphaltpavementdistressdata
collectionappliedonthemainrunwayofAmílcarCabralinternationalairport,locatedatSal
IslandinCapeVerde.Thetwomethodsusedfortestingweretraditionalvisualinspection
(on-foot)andanindirectmethodusingavehicleequippedwithimagecaptureandrecording,lasers
andgeolocationdevices(in-vehicleinspection).Theaimofthisresearchistocontributetothe
validation oftheproposedlow-costin-vehicle pavementdistressinspectionsystemwith
semiautomatic data processing in order to be considered in the implementation ofthe pavement
conditionassessmentcomponentofairportpavementmanagementsystems(APMS).Thisisa
particularlyimportantcomponentasfromthecollecteddistressdataitispossibletoassessthe
conditionofthepavementsanddefineinterventionstrategies.Validationoftheindirectdata
collectionmethodisevaluatedbystatisticalcomparisonofthecollecteddistressdataand
pavementconditionindex(PCI)obtainedfrombothmethods.Statisticallynon-significant
differencesbetweentheresultsetsvalidatetheproposedindirectmethod,howevertheanalysis
evidenced two aspects that need improvement in the proposed system, namely the quality ofthe
capturedimagestoidentifydistresseswithlowerseverity levelandinspectortrainingforproper
allocationof severitylevels duringimage analysis.Thisresults insignificantadvantages
consideringthatthetotalamountoftherunwaypavementareaisinspected.Inspectiontimeis
reducedanddatacollectioncostcanbereduced.ProcessingandresultsvisualizationonGIS
environment allows revaluation ofthe dataset on the in-vehicle method. Data interpretation and
measurementsqualitycontrolbecomessimplerandfaster.
©2020TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCC
BY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).
1.Introduction
ThedevelopmentandimplementationofanAPMSmakesitpossibletomanageassetsinapracticalandsustainable manner, assisting managementpersonnel indecision makingthrough prioritysetting,cost quantificationand activity scheduling,resultinginthedevelopmentofeconomicallyviablestrategiesforpavementmaintenance[1–3].Fig.1presents themaincomponentsthatareconsideredontheimplementationofanAPMStooperateatthenetworkorprojectlevel[4].
*Correspondingauthor.
E-mailaddresses:bsantos@ubi.pt(B.Santos),galmeida@ubi.pt(P.G. Almeida),ianca.feitosa@ubi.pt(I.Feitosa),debora.lima@asa.cv(D.Lima).
https://doi.org/10.1016/j.cscm.2020.e00419
2214-5095/©2020TheAuthors.PublishedbyElsevierLtd.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/ licenses/by-nc-nd/4.0/).
ContentslistsavailableatScienceDirect
Case
Studies
in
Construction
Materials
Informationrelatedtothecharacteristicsoftheairportpavementnetwork,asconstruction,maintenance,trafficand condition data are collected in the network inventory component. In condition assessment, condition data, defined accordinglytotheevaluationmethodandsurveyfrequencyadoptedbythesupervisingagency,isusedtodeterminethe pavementcondition.Dataisorganizedandstoredinadatabaseforuseindataanalysis,whereevaluationandprioritization oftherehabilitationandbudgetneedsovertheanalysisperiod isperformed.Finally,themain outputs(resultsofthe analyses)areorganizedindifferentformatsforconsiderationofengineersandmanagers.
Thefeedbackloop,oftenoverlooked,establishesaprocesswhereactualperformanceandcostdataareinputtedbackinto themodelsusedonpavementmanagementanalyses[4].
Fromtheabove,itispossibletoconcludethatefficientairportmanagementneedsasignificantamountofdataregarding theconditionofairportpavements.Qualityofthecollecteddataisthereforeessentialtoidentifydeficienciesandproperly diagnose pavement condition. This allows to detect current and future maintenance needs predicting the effects of interventionstrategyoverpavements’lifetime[6,7].
Dataisgenerallycollectedbyvisualinspectionofairportpavements.However,visualcollectionofsurfacedistressesis becoming increasingly challenging for airport maintenance teams as operational constraints limit access time for inspectionsonhighprioritypavements.Moreoverreductionsin fundingandoperationalstaffresourcesleadtovisual inspectionconstraints[8].Consequently,choosingthedatacollectionmethodthatwillprovidethenecessaryinformationto evaluatetheconditionofpavementsisvitalforAPMSsuccess.
Datacollectionbyvisualinspectionhasevolvedfromthetraditionalvisualconditionsurveyperformedonfoot,where distresstype,severity,andquantityareidentifiedandrecordedinpaperform,totheuseofmultifunctionalvehicles.Data collectionsystems installed onthese vehicles arecomposed by one or moreacquisition devices and post-processing applicationsforsemiautomaticorautomateddataextractionprocedures,basedoncomputervisionandimageprocessing algorithms[9].Devicesaregenerallyusedbothonroadandairportenvironments.Amongthemostcommon,thosestanding outareGNSS(locationreferencing)and videologgingmodules(forpavementsand360 imagery),lasercrackand rut measurementsystem(cracks,transverseprofileandrutting),highresolutionodometer(lineardistance),laserprofilometer (roughness– IRI,macrotextureand longitudinalprofile)andlandmobilelightdetectionand ranging(LiDAR)systems. Severalreferencedauthorstested,discussedanduseddataobtainedbythesemodulesandsystemsonairportpavement inspection[8,10–12].
Alsoimportanttonoticeisthelatestdevelopmentsinroadandairportassetinspectiontechnology,thatfocusontheuse ofunmannedaerialvehicle(UAVs-drones)includingLiDARtechnology[13–15].
Inthis context,anin-vehiclesystemwithimagecapture andrecording,lasersand geolocationdevices thatcanbe classifiedas alow-cost, mobilemultifunctionsystemwithsemiautomatic dataprocessingisproposed. Thisin-vehicle approachhasbeendevelopedoverthelast4yearsatthedepartmentofcivilengineeringandarchitectureoftheUniversityof BeiraInterior(UBI)[5,16–22],aimingtoimprovetheefficiencyofdatacollectionatlocalandsmallinternationalairports.So, thisstudyaimstoverifytheeffectivenessoftheindirectin-vehicleinspectionmethod,that‘àpriori’seemstobeamore practicalmethodologytouseasitrequireslesstimeontherunwayandfewerresourcesthanon-footinspection(withoutthe aidofspecializedtechnicalapparatus).
Forthispurpose,resultsobtainedfromatraditionalon-footpavementsurfacedistressinspectionandtheproposed in-vehicleinspectionsystemdevelopedatUBIarecompared.Acomparativestatisticalanalysiswasperformedontheresultsof pavementsurfacedistressdata(type,severitylevelsanddensities)collectedatexactlythesamesamplingunitsofthemain runwayofAmílcarCabralinternationalairport(ICAO:GVAC),locatedatSalIslandinCapeVerde.
Thepavementconditionindex(PCI)obtainedbyLima(2016)[21]forbothmethodsisalsocompared.PCIforairport pavementswasdevelopedbytheUSArmyCorpsofEngineersthroughfundingprovidedbytheU.S.AirForce.Thisindex describesthestructuralintegrityandoperationalconditionofpavementsanditsdeterminationisbasedontheidentification andassessmentoftheseveritylevelof17distressesobservedonpavementssurfaces[23].Itrangesfrom0(worstcondition) to100(bestcondition)anditsratingscalecanberelatedtomaintenanceandrehabilitationactivities[2,3,12,20].
Thispaperisorganizedin5sections:thefirstsectionpresentstheconceptofAPMSandtheimportanceofchoosingthe airportpavementinspectiontechnologyforthesuccessofthepavementconditionassessment;thefollowingthreesections describethestatisticalapproachadoptedtovalidatetheindirectpavementdatacollectionmethodtested,presentthecase studyandthestatisticalanalysisresults;finally,thelastsectionconcludeshighlightingsomereflectionsoftheauthorsbased ontheresultsobtainedandpresentingtopicsforfurtherdiscussion.
2.Methodology
Tovalidatetheproposedindirectdatacollectionmethod(in-vehicle),thecollecteddistressdataandPCIvaluesobtained throughonfootandin-vehicleinspectionwerestatisticallycompared.
Statisticalcomparisonofgroupsdependsonhowthevariablesunderstudyaremeasured(qualitative/quantitative),its qualityandquantity,typeofsampling(probabilisticorrandom/non-probabilisticornon-random),relationshipbetween samples(independent/paired),thesamples’normalitydistribution(normal/non-normal)andthenumberofgroupstobe compared.
Inthisstudy,variablesreferringtopavementdistresstype,severityanddensitywereevaluatedonsamplesfromthe GVACrunway01 19(studypopulation).SamplingwasdefinedaccordingtoASTMD5340-12[23].Thisstandardadopts systematicrandomsamplingforselectionofsampleunitstobeinspectedforpavementconditionassessment.
Sincethestudycomparesbothpavementinspectionmethods,pairedsamplesconstitutedbythesamesubjectswere considered(comparisonoftwo-pairedgroups),sopavementdistressdatawascollectedusingthetwoinspectionmethods ontheexactlysamesamplingunitsoftherunway.
Statisticalcomparisonofthesamplesisthenevaluatedbystatisticaltests.Testchoicedependsfundamentallyonthe number ofgroupstobecompared andonnormalityofthedata(variables),assessedbytheKolmogorov-Smirnovand Shapiro-Wilkstests.
Variables with normal distribution are statistically analysed by parametric t-test, while those with non-normal distributionareevaluatedusingtheWilcoxonnon-parametrictest[24].t-testisusedtodeterminetheoccouranceofa significantdifferencebetweenthemeansoftwo-pairedgroups.OntheWilcoxontest,differencesbetweensetsofpairsare calculatedandanalyzedinordertoestablishifthetwogroupsarestatisticallysignificantlydifferent[24].
3.Casestudy
Thecasestudywasdevelopedatrunway01 19oftheAmílcarCabralinternationalairport(ICAO:GVAC),locatedabout threekilometressouthofthecityofEspargos,atSalIsland,CapeVerde.CapeVerde'sairportsystemconsistsofsevenairport infrastructures.ItisthemaindriveroftheCapeVerdeaneconomyasthecountrymainactivityistourism.Inaddition,the geographyoftheterritory,anarchipelago,contributestotheincreasedrelevanceofairtransport.Issuessuchasaccessibility betweenislandsandtoothercountriesbecomesacentraldemandfortheeconomicdevelopmentandsocialactivityof citizens.
Fordistressinspection,theairportpavementnetworkisdividedinto4brancheswitha totalof16sections.GVAC's runwayis45mwideby3000mlong,with7.5mwideshoulders,comprising3oftheaforementionedsectionswith270 samplingunits.Theentiresurfaceoftherunwaypavementwasinspectedbythein-vehiclesystem(270samplingunits)and 43wereinspectedonfoot.This43samplingunitsconstitutestheminimumnumbertobeinspectedaccordingtoASTM D5340-12forPCIevaluation[20].Table1summarizesthesegmentationandadoptedcodinginformation.
Table1
CodingandsegmentationofGVACrunway01–19.
Pavement Code Description/Comments
Pavementnetwork–Airport SID GVACpavementnetwork
Pavementbranches R01 Mainrunway(01 19)
TWY Taxiway
BER Shoulders
APR Apron
R01pavementsections A 900mmeasuredfromend01
(touchdownarea)
B 1200mlocatedinthecentralpartofthe runway
C 900mmeasuredfromend19
(touchdownarea) Pavementsamplingunits–totalrunwayarea
(segmentationaccordingtoASTMD5340-12sampleareacriteria)
A:81samplingunits 500m2
samplingunits(100mby5m) B:108samplingunits In-vehiclevisualinspection
C:81samplingunits Pavementsamplingunits–sampleareaforinspection
(minimumareatoinspectaccordingtoASTMD5340-12)
A:14samplingunits 500m2samplingunits(100mby5m)
B:15samplingunits Traditionalandin-vehiclevisual inspection
Itshouldbenotedthatdespitehavingperformedatotalrunwayin-vehicleinspection,onlythecommon43sampling unitswereusedonthecomparativestatisticalanalysis.Fig.2presentsthelocationofthe3pavementsectionsstudied.
Onthein-vehicleinspectionthefollowingpavementdistressesaredirectlyidentifiedondaytimeimages:alligator cracking;bleeding;blockcracking;jet-blasterosion;jointreflectioncracking;longitudinalandtransversecracking;oil spillage;patching andutilitycut patch;polishedaggregate; ravelling; shoving;slippage crackingandweathering. Laser-lineprojectionimagescapturedatnightallowstheidentificationofcorrugation,rutting,swellinganddepression distresses.These 17 distressesare the ones consideredin ASTMD5340-12 standardfor PCIcalculation.A detailed description,aswellasthedefinitionofseveritylevels(low,mediumandhigh)andhowtomeasuredistressescanbe consultedin[23,25].
Thus,inordertocollectallthedistressdata,twoin-vehicleinspectionswerecarriedout,oneduringthedaytimeandthe otherat night.Takingintoaccounttheaccessrestrictionstotherunway duetoaircrafttraffic,allinformation forthe completerunwaywascollectedwithin3days,whiletheon-footinspectiontook3weeksforonly43samplingunits.
Uponcompletionofbothinspections,ageoreferenceddatasetofimageswasobtainedandprocessedforvisualizationon ageographicinformationsystem(GIS).Imageswerecollectedtoformacontinuousstreamwithenoughresolutiontoallowa spatialresolutionoflessthan0.003mperpixelatthecentreoftheimageandreducingspatialresolutiontowardsthe marginsduetogeometricdistortionassociatedwiththeopticalsystem.Latertheimageswerecorrelatedwithpositioning datagatheredonadualbandGPSreceiverandpost-processedonaPPP(precisepointpositioning)strategyusingonlinefree processingresources.Thisallowedforlocationerrorsfarbelow0.05mwhilemaintainingequipmenttoaminimum.Image and positioning data correlation quality was verified using control points on the ground. Image georeferencing and orientationwasperformedusinganin-housedevelopedsoftwarepackage.
Imagevisualizationoccursonacontinuouscoverageoverimposedreferencefeaturesoftherunwaytopographicsurvey. Thisvisualizationallowedidentificationofsurfacepavementdistressesandseveritylevelaswellasmeasurementofareas andlengthofinfluence.Theon-screendigitisationprocessofdistresseslengthandareafromcapturedimagesallowsfora vertexplacementaccuracythatis,onaverage,of2pixels,correspondingto0.006m.Measurementsperformedon-footusing measuringwheel presentsa sub-decimetricaccuracy forlineardistresses.Surfacedistressareaiscalculatedusingthe productoftwoapproximatelyorthogonallinearmeasurements.Consequently,measurementofareasusingimages,basedon careful,on-screenrastertovectordigitisation,aremoreaccuratethaninsitumeasurementsusingameasuringwheel.
Theinspectionsidentifiedthepresenceof6typesofpavementdistressesonGVACrunway.Acomparativeanalysisofthe dataobtainedbythetwoinspectionapproachesshowsadiscrepancyregardingtheamountofidentifieddistresstypesand densitybyseveritylevel(seeTable2).
While6distresseswereidentifiedbyon-footinspectiononly4wereregisteredusingthein-vehiclesystem.Thetwo distressesnotidentifiedduringthein-vehicleinspectionpresentessentiallyalowdensityandseveritylevelontheon-foot inspection.Inthecaseofdepressions,onlyalowlevelofseveritywasidentifiedon-foot,whichisthoughttohaveimpaired identification usingthein-vehicle system asit is difficulttodetectthis levelof severityonthecapturedimages. For longitudinal and transverse cracking at low severity level, the same situation occurs, however it is possible that a misidentificationoccurredathighseveritylevel,probablyonareaspresentingalligatorcracking.Previoustestsofthe in-vehiclesystemonroadenvironmentsuccessfullyidentifiedcrackingsituationsatahighseveritylevel[22].
Severityleveldensitydistributionshowsabiastowardsthehigherseverityclassondistressescollectedbythein-vehicle system.However,thesumofcommondistressesdensitiesforbothmethodsissimilar.Toassesswhetherthesedifferences arestatisticallysignificant,astatisticalcomparisonofthetwogroupsofsampleswasperformed.
It is worthtopoint out thatdata processingof theinformation gatheredduring thetwoinspection methodswas performedbythesameinspector.
Subsequentlyandconsideringthe6(on-footinspection)and4(in-vehicleinspection)identifiedstresses,PCIvalueswere calculatedperinspectedsamplingunitandpavementsectionforbothmethods(seeTable3andFig.3).Valuesobtainedby bothapproachesaresimilar,withmostoftheresultsshowingPCIvaluesintherange25–40,reflectingtheneedforashort terminterventionontherunwaypavement.
Fig.3presentsthescatterplotofPCIvaluespairedbysamplingunitandthecorrespondinglinearregressionmodel.The linearmodelisagoodfitforthedata,asdifferencesbetweenobservationsandthepredictedvaluesaresmallandunbiased (fittedvaluesarenotsystematicallytoohighnortoolowintheobservationspace).Theseconclusionsareemphasizedbya R2-valueof0.8993andastandarderroroftheestimatesof3.6386.
Consideringanoptimalmodelwherein-vehiclePCIwouldbethesameason-footdeterminedPCI,representedonFig.3 asareferenceline,itisclearthatlowerPCIvaluesareoverestimatedonthein-vehicleapproachandhighervaluesare underestimated,withapivotingpointaroundPCI=32.
4.Statisticalanalysisanddiscussionofresults
Densityvaluesof4pavementdistressesthatwereidentifiedonGVAC’srunwaybyon-footandproposedin-vehicle inspectionsystemwereusedtostatisticallycomparethetwoinspectionapproaches.ThiswasalsoperformedforthePCI valuesobtainedfromthoseidentifieddistresses.
Forthispurpose,threeanalyseswereimplemented:descriptivestatisticalanalysisconsideringdispersionandcentral tendencymeasures,anormalityanalysistodeterminethecomparativestatisticaltesttobeappliedduringthecomparative study(parametricornonparametric)andastatisticalcomparativeanalysis.
OnthenormalityanalysistheKolmogorov-SmirnovandShapiro-Wilktestswereadoptedandappliedtodensityvalues bytypeofdistressandlevelofseverity(low,mediumandhigh).TheobtainedP-valuesindicatethatbothdatasets(on-foot andin-vehicle)essentiallypresentanon-normaldistribution(p-value<0.05).Theexceptionbeingweatheringinsurface wear(highseveritylevel)onin-vehicleinspectionthatexhibitsanormaldistributionaccordingtotheKolmogorov-Smirnov test.Thisresultcanberelatedtothehighernumberofidentifiedcaseswhencomparedtotheother3distresses(larger sample size),suggestingan influenceontheresultof thenormalitytest (seeTables2and 4).Duetothenon-normal
Table2
Identifiedpavementdistressesanddensitiesbyseveritylevelandtypeofinspection.
Inspectiontypeandidentifieddistresses Densitybyseveritylevel(%)*
On-footinspection Low Medium High
Alligatorcracking 2.00 1.83 0.07
Patchingandutilitycutpatch 2.18 1.02 1.92
Ravelling 0.70 2.66 11.13
Weatheringinsurfacewear-densemixasphalt 4.78 7.82 16.00
Depressions 0.53 0.00 0.00
Longitudinalandtransversecracking 0.24 0.00 0.16
Sum(alldistressesidentified):53.04 10.43 13.33 29.28
Sum(distressesidentifiedonbothmethods):52.11 9.66 13.33 29.12
In-vehicleinspection Low Medium High
Alligatorcracking 1.09 2.23 0.00
Patchingandutilitycutpatch 0.10 2.35 1.15
Ravelling 0.00 2.75 12.60
Weatheringinsurfacewear-densemixasphalt 1.96 3.26 23.39
Sum(identifieddistresses):50.88 3.15 10.59 37.14
Table3
PCIvaluesobtainedforeachsampleunitinspectedandrunwaysection(adaptedfrom[21]).
Section Typeofinspection PCIpersamplingunit PCIpersection
A On-foot 36,38,41,43,52,20,52,27,34,46,24,11,60,46 38 In-vehicle 30,40,45,40,52,22,46,30,37,46,26,15,54,48 38 B On-foot 22,32,48,25,22,15,17,18,27,56,38,53,31,35,20 30 In-vehicle 26,26,41,27,28,24,18,16,30,59,38,47,31,29,25 31 C On-foot 44,27,32,36,35,20,35,63,32,43,59,25,30,26 36 In-vehicle 47,23,32,32,29,19,35,53,32,39,54,29,24,28 34
distributionofthedistressdensitydata,thenon-parametricWilcoxontestwasadoptedforcomparativeanalysisofboth groups.
ConsideringPCIvalues,theShapiro-Wilknormalitytestindicatesanormaldistributionofdata(P-value>0.05)forboth on-footandin-vehicleinspection.However,accordingtotheKolmogorov-Smirnovtest,PCIvaluesobtainedfromin-vehicle inspectiondatadonotfollowanormaldistribution.Giventheseresults,aPCIstatisticalcomparisonwithparametrict-test (statisticalmean)andnonparametricWilcoxon test wasperformed. Anf-testwas alsoperformed tocompare sample variance.
ResultsofthenormalitytestsandthestatisticalcomparisonanalysisarepresentedinTables4and5.
Resultsshowthatpatchingandutilitycutpatchdistressidentifiedatlowandmediumseveritylevelswerenotvalidated. Inthesecases,P-valuesarelessthan0.05,meaningthatthesamplesarenon-comparable.Thesameoccursforweatheringin surfacewearathighseveritylevelanditcanberelatedtothediscrepancyofthenormalityresultsobservedforthisparticular case.
Thenon-validationofthesecasescanberelatedtoclassificationdifficultiesofweatheringinsurfacewearataspecific severityleveloncollectedimages,aswellasotherpavementdistressespresentinglowseveritylevel.
Fig.3.PairedvaluesofPCIobtainedinthetwoinspections(on-footandin-vehicle)andlineartrendline.
Table4
Normalitytestsresults.
Test Kolmogorov-SmirnovP-value Shapiro-WilkP-value
InspectionbyfootSeveritylevel Low Medium High Low Medium High
Alligatorcracking 0.000 0.000 0.000 0.000 0.000 0.000
Patchingandutilitycutpatch 0.000 0.000 0.000 0.000 0.000 0.000
Ravelling 0.000 0.000 0.000 0.000 0.000 0.000
Weatheringinsurfacewear-densemixasphalt 0.000 0.000 0.011 0.000 0.000 0.000
PCI 0.200 0.340
In-vehicleinspectionSeveritylevel Low Medium High Low Medium High
Alligatorcracking 0.000 0.000 * 0.000 0.000 *
Patchingandutilitycutpatch 0.000 0.000 0.000 0.000 0.000 0.000
Ravelling * 0.000 0.000 * 0.000 0.000
Weatheringinsurfacewear-densemixasphalt 0.000 0.000 0.200 0.000 0.000 0.002
PCI 0.008 0.097
*
Theremainingdistresses-severitypairsandPCIvalueswerevalidatedbytheWilcoxontest.t-testalsovalidatedPCI values.
Basedontheresultsoft-test(statisticalmean)andf-test(variance),itcanbeinferredthattheuseoftheproposed in-vehiclesystemtoinspect pavementsurfacedistressesdidnotsignificantlyinfluencePCIresults,sincethestatistical average(t-test)ofbothsamples,aswellasitsvariance(f-test)producedP-valuesgreaterthan0.05,validatingtheproposed in-vehiclesystem.
5.Conclusions
ThecomparativestatisticalanalysisonPCIvaluesvalidatestheinnovativelow-costmethodproposedforcarryingout inspectionsusinganequippedvehicle.Foramoreconsistentvalidation,distressesidentifiedinbothapproacheswerealso meticulouslyanalysedbylevelofseverity.
Dataanalysisevidencedabetteridentificationofdistressesandseveritylevelsbydirectvisualization,inparticularforlow severitylevelsandweatheringinsurfaceweardistress.
Comparingthetypeofdistressesidentifiedbyeachapproach,it’sclearthatthein-vehiclemethodcouldnotassesstwo typesofdistressesthatweresurveyedon-footatalowseveritylevelandlowdensityvalues.Still,bycomparingdistresses allocationbyseveritylevel,it’sclearthatatendencytoclassifydistressesathigherseverityleveloccursonthein-vehicle system.
Theseresults,supportedbythestatisticalcomparisonanalysis,donotcompromisetheproposedmethod,sinceitallowed toidentifyimprovementaspectsintheproposedin-vehiclesystemandinthesemi-automaticprocessingofimages.
Theauthorsconsiderthat, byimprovingthein-vehicle system,thedifficulties encounteredwhileidentifyingsome distresseswillbesurpassed.Thisimprovementcanbeachievedthroughtheuseofhigherimagequalitycamerasandthe additionofasecondcameraplacedtocapturepavementroughnessandsmalldeformations.Itisfurtherconsideredthat inspectors trainingonimage analysis,especially regarding distressseveritylevelallocation, aswellas theinspectors’ accumulatedexperiencecouldpositivelyinfluenceresultsofthecomparativestatisticalanalysis.
Bycomparingbothinspectionapproachesitwasnoticedthattheproposedin-vehicleprototypewithimage,laserand GNSSdeviceshasasthemainadvantagethereductionofthetimeneededtocollectpavementdistressinformation,as traditional(on-foot)surveysaretimeconsuming,becomingmoreexpensiveforlargerairportswithtrafficconstraints.
Thein-vehicleapproachalsoallowstocontinuouslysurveytheconditionoftheentirepavementsurfaceandtoprocessall theinformationattheofficevisualizingdataonaGISenvironment,allowingre-evaluationandconfirmation,ifnecessary,of thepavementdistressdataobtainedunderthesameconditionsasthoseverifiedatthetimeofinspection(imagearchive). WiththisapproachitisalsopossibletocomparetheGISimageandalphanumericdatafromseveralinspectionscarried outoverthepavementlifetime,supportingtrendstudiesofpavementcondition.
DeclarationofCompetingInterest
Theauthorsreportnodeclarationsofinterest. Acknowledgements
TheauthorsacknowledgeUniversityofBeiraInterior,CERIS-CivilEngineeringResearchandInnovationforSustainability (ECI/04625),GEOBIOTEC-GeoBioSciences,GeoTechnologiesandGeoEngineering(GEO/04035) andASASA,CapeVerde AirportsandAirSafetyforsupportandfundingofthisstudy.
References
[1]M.Y.Shahin,PavementManagementforAirports,Roads,andParkingLots,secondedition,SpringerUS,2005,doi:http://dx.doi.org/10.1007/b101538. [2]FederalAviationAdministration(FAA),AC150/5380-7B,AirportPavementManagementProgram(PMP),(2014).https://doi.org/10.1004/AC150/5380-7B. [3]T.Jensen,C.M.Lang,C.MotzHagerstownRegionalAirportRichardTucker,P.P.Hochstetler,N.AmericaLauraMcKee,J.BarryBarker,D.H.Butler,J.M.
Crites,A.K.Kanafani,C.Professor,M.R.Morris,T.L.Rosser,H.G.Schwartz,B.A.Scott,K.T.Steudle,M.Dot,L.W.DouglasStotlar,C.MichaelWalton, CommonAirportPavementMaintenancePracticesASynthesisofAirportPractice,(2011).
Table5
Wilcoxontest,t-testandf-testresults.
Variable WilcoxontestP-value t-testP-value f-testP-value
Severitylevel Low Medium High Notapplicable Notapplicable
Alligatorcracking 0.068 0.657 0.317
Patchingandutilitycutpatch 0.012 0.011 0.076
Ravelling 0.068 0.861 0.640
Weatheringinsurfacewear-densemixasphalt 0.114 0.122 0.008
[4]W.DOT,WashingtonAirportPavementManagementManualWashington,(2019).
[5]D.Lima,B.Santos,P.G.Almeida,ProposalofanairportpavementmaintenancemanagementsystemforCapeVerde,STARTCON19–Int.Dr.Students Conf.+LabWork.Civ.Eng.,KnEEngineering,Covilhã,2020,pp.49–60,doi:http://dx.doi.org/10.18502/keg.v5i5.6917.
[6]L.Pierce,G.McGovern,K.Zimmerman,PracticalGuideforQualityManagementofPavementConditionDataCollection,(2013),pp.170.https://www. fhwa.dot.gov/pavement/management/qm/data_qm_guide.pdf.
[7]S.Saliminejad,N.G.Gharaibeh,Impactoferrorinpavementconditiondataontheoutputofnetwork-levelpavementmanagementsystems,Transp. Res.Rec.(2013)110–119,doi:http://dx.doi.org/10.3141/2366-13.
[8]K.Keegan,Katherine,Jung,InnovativeapproachtoairfieldpavementinspectionsanddistressidentificationatOaklandInternationalAirport,Int.Conf. Manag.PavementAssets,Washington,DC,2015.https://vtechworks.lib.vt.edu/bitstream/handle/10919/56397/ICMPA9-000070.PDF?
sequence=2&isAllowed=y.
[9]A.Ragnoli,M.R.deBlasiis,A.DiBenedetto,Pavementdistressdetectionmethods:areview,Infrastructures3(2018)1–19,doi:http://dx.doi.org/ 10.3390/infrastructures3040058.
[10]W.Chen,J.Yuan,M.Li,ApplicationofGIS/GPSinShanghaiairportpavementmanagementsystem,ProcediaEng.(2012)2322–2326,doi:http://dx.doi. org/10.1016/j.proeng.2012.01.308.
[11]M.Barbarella,M.R.DeBlasiis,M.Fiani,Terrestriallaserscannerfortheanalysisofairportpavementgeometry,Int.J.PavementEng.20(2019)466–480, doi:http://dx.doi.org/10.1080/10298436.2017.1309194.
[12]P.DiMascio,L.Moretti,Implementationofapavementmanagementsystemformaintenanceandrehabilitationofairportsurfaces,CaseStud.Constr. Mater.(2019),doi:http://dx.doi.org/10.1016/j.cscm.2019.e00251.
[13]C.Brooks,R.Dobson,D.Banach,T.Oommen,K.Zhang,A.Mukherjee,T.Havens,T.Ahlborn,R.Escobar-Wolf,C.Bhat,S.Zhao,Q.Lyu,M.D.of Transportation,ImplementationofUnmannedAerialVehicles(UAVs)forAssessmentofTransportationInfrastructure-PhaseII164p,,(2018).https:// www.michigan.gov/documents/mdot/SPR-1674_FinalReport_revised_631648_7.pdf%0Ahttps://trid.trb.org/view/1542464.
[14]P.Complete,B.Task,R.Inspections,T.Transfer,TheApplicationofUnmannedAerialSystemsinSurfaceTransportation,(2019).
[15]L.Jiang,Y.Xie,T.Ren,Adeepneuralnetworksapproachforpixel-levelrunwaypavementcracksegmentationusingdrone-capturedimages,Transp. Res.Board99thAnnu.Meet.(2020).
[16]L.Maganinho,DevelopmentofaDatabaseforRoadPavementConditionUsingGPS,VideoImageandGIS(inPortuguese),UniversityofBeiraInterior, 2013.
[17]A.Nogueira,EvaluationofRoadPavementsRutswithLaserScanning(inPortuguese),UniversityofBeiraInterior,2015. [18]D.Lima,AirportPavementManagementSystemforCapeVerde(inPortuguese),UniversityofBeiraInterior,2016.
[19]A.Domingos,PavementConditionIndexDeterminationandInterpretationforAirportPavementEvaluation(inPortuguese),UniversityofBeira Interior,2017.
[20]A.Domingos,B.Santos,P.G.Almeida,Pavementconditionindexdeterminationandinterpretationforairportpavementevaluation,Int.Congr.Eng., ICEUBI2017,Covilhã,2017.
[21]D.Lima,B.Santos,P.G.Almeida,MethodologytoassessairportpavementconditionusingGPS,laser,videoimageandGIS,PavementAssetManag. -Proc.WorldConf.PavementAssetManag.,WCPAM2017,2019,pp.301–307.
[22]B.Santos,P.G.Almeida,L.Maganinho,DatacollectionmethodologytoassessroadpavementconditionusingGNSS,videoimageandGIS,IOPConf.Ser. Mater.Sci.Eng.,InstituteofPhysicsPublishing,,2019,doi:http://dx.doi.org/10.1088/1757-899X/603/4/042083.
[23]AmericanSocietyforTestingandMaterials,D5340-12,StandardTestMethodforAirportPavementConditionIndexSurveys,Reproduction,(2012),pp. 1–54,doi:http://dx.doi.org/10.1520/D5340-12.2.
[24]J.Marôco,AnáliseEstatísticacomoSPSSStatistics,6a
edition,(2014) Portugal.