A survey on wireless multimedia sensor networks.docx

A survey on wireless multimedia sensor networks.docx

《A survey on wireless multimedia sensor networks.docx》由会员分享，可在线阅读，更多相关《A survey on wireless multimedia sensor networks.docx（130页珍藏版）》请在冰豆网上搜索。

Asurveyonwirelessmultimediasensornetworks

Asurveyonwirelessmultimediasensornetworks

IanF.Akyildiz*,TommasoMelodia,KaushikR.Chowdhury

BroadbandandWirelessNetworkingLaboratory,SchoolofElectricalandComputerEngineering,GeorgiaInstituteofTechnology,

Atlanta,GA30332,UnitedStates

Received11March2006;receivedinrevisedform6August2006;accepted5October2006

Availableonline2November2006

Abstract

Theavailabilityoflow-costhardwaresuchasCMOScamerasandmicrophoneshasfosteredthedevelopmentofWire-

lessMultimediaSensorNetworks（WMSNs）,i.e.,networksofwirelesslyinterconnecteddevicesthatareabletoubiqui-

touslyretrievemultimediacontentsuchasvideoandaudiostreams,stillimages,andscalarsensordatafromthe

environment.Inthispaper,thestateoftheartinalgorithms,protocols,andhardwareforwirelessmultimediasensornet-

worksissurveyed,andopenresearchissuesarediscussedindetail.ArchitecturesforWMSNsareexplored,alongwith

theiradvantagesanddrawbacks.Currentlyoff-the-shelfhardwareaswellasavailableresearchprototypesforWMSNs

arelistedandclassified.Existingsolutionsandopenresearchissuesattheapplication,transport,network,link,andphys-

icallayersofthecommunicationprotocolstackareinvestigated,alongwithpossiblecross-layersynergiesand

optimizations.

_2006ElsevierB.V.Allrightsreserved.

Keywords:

Wirelesssensornetworks;Multimediacommunications;Distributedsmartcameras;Videosensornetworks;Energy-aware

protocoldesign;Cross-layerprotocoldesign;Qualityofservice

1.Introduction

Wirelesssensornetworks（WSN）[22]havedrawn

theattentionoftheresearchcommunityinthelast

fewyears,drivenbyawealthoftheoreticaland

practicalchallenges.Thisgrowinginterestcanbe

largelyattributedtonewapplicationsenabledby

large-scalenetworksofsmalldevicescapableofhar-

vestinginformationfromthephysicalenvironment,

performingsimpleprocessingontheextracteddata

andtransmittingittoremotelocations.Significant

resultsinthisareaoverthelastfewyearshaveush-

eredinasurgeofcivilandmilitaryapplications.As

oftoday,mostdeployedwirelesssensornetworks

measurescalarphysicalphenomenaliketempera-

ture,pressure,humidity,orlocationofobjects.In

general,mostoftheapplicationshavelowband-

widthdemands,andareusuallydelaytolerant.

Morerecently,theavailabilityofinexpensive

hardwaresuchasCMOScamerasandmicrophones

thatareabletoubiquitouslycapturemultimedia

contentfromtheenvironmenthasfosteredthe

1389-1286/$-seefrontmatter_2006ElsevierB.V.Allrightsreserved.

doi:

10.1016/net.2006.10.002

*

Correspondingauthor.Tel.:

+14048945141;fax:

+1404894

7883.

E-mailaddresses:

ian@ece.gatech.edu（I.F.Akyildiz）,tomma-

so@ece.gatech.edu（T.Melodia）,kaushikc@ece.gatech.edu

（K.R.Chowdhury）.

ComputerNetworks51（2007）921–960

ofWirelessMultimediaSensorNet-

works（WMSNs）[54,90],i.e.,networksofwirelessly

interconnecteddevicesthatallowretrievingvideo

andaudiostreams,stillimages,andscalarsensor

data.Withrapidimprovementsandminiaturization

inhardware,asinglesensordevicecanbeequipped

withaudioandvisualinformationcollectionmod-

ules.Asanexample,theCyclopsimagecapturing

andinferencemodule[103],isdesignedforextre-

melylight-weightimagingandcanbeinterfaced

withahostmotesuchasCrossbow’sMICA2[4]

orMICAz[5].Inadditiontotheabilitytoretrieve

multimediadata,WMSNswillalsobeabletostore,

processinreal-time,correlateandfusemultimedia

dataoriginatedfromheterogeneoussources.

Wirelessmultimediasensornetworkswillnot

onlyenhanceexistingsensornetworkapplications

suchastracking,homeautomation,andenviron-

mentalmonitoring,buttheywillalsoenableseveral

newapplicationssuchas:

•Multimediasurveillancesensornetworks.Wireless

videosensornetworkswillbecomposedofinter-

connected,battery-poweredminiaturevideo

cameras,eachpackagedwithalow-powerwire-

lesstransceiverthatiscapableofprocessing,

sending,andreceivingdata.Videoandaudio

sensorswillbeusedtoenhanceandcomplement

existingsurveillancesystemsagainstcrimeand

terroristattacks.Large-scalenetworksofvideo

sensorscanextendtheabilityoflawenforcement

agenciestomonitorareas,publicevents,private

propertiesandborders.

•Storageofpotentiallyrelevantactivities.Multime-

diasensorscouldinferandrecordpotentiallyrel-

evantactivities（thefts,caraccidents,traffic

violations）,andmakevideo/audiostreamsor

reportsavailableforfuturequery.

•Trafficavoidance,enforcementandcontrolsys-

tems.Itwillbepossibletomonitorcartrafficin

bigcitiesorhighwaysanddeployservicesthat

offertrafficroutingadvicetoavoidcongestion.

Inaddition,smartparkingadvicesystemsbased

onWMSNs[29]willallowmonitoringavailable

parkingspacesandprovidedriverswithauto-

matedparkingadvice,thusimprovingmobility

inurbanareas.Moreover,multimediasensors

maymonitortheflowofvehiculartrafficon

highwaysandretrieveaggregateinformation

suchasaveragespeedandnumberofcars.Sen-

sorscouldalsodetectviolationsandtransmit

videostreamstolawenforcementagencies

toidentifytheviolator,orbufferimagesand

streamsincaseofaccidentsforsubsequentacci-

dentsceneanalysis.

•Advancedhealthcaredelivery.Telemedicinesen-

sornetworks[59]canbeintegratedwith3Gmul-

timedianetworkstoprovideubiquitoushealth

careservices.Patientswillcarrymedicalsensors

tomonitorparameterssuchasbodytemperature,

bloodpressure,pulseoximetry,ECG,breathing

activity.Furthermore,remotemedicalcenters

willperformadvancedremotemonitoringof

theirpatientsviavideoandaudiosensors,loca-

tionsensors,motionoractivitysensors,which

canalsobeembeddedinwristdevices[59].

•Automatedassistancefortheelderlyandfamily

monitors.Multimediasensornetworkscanbe

usedtomonitorandstudythebehaviorofelderly

peopleasameanstoidentifythecausesof

illnessesthataffectthemsuchasdementia[106].

Networksofwearableorvideoandaudiosensors

caninferemergencysituationsandimmediately

connectelderlypatientswithremoteassistance

servicesorwithrelatives.

•Environmentalmonitoring.Severalprojectson

habitatmonitoringthatuseacousticandvideo

feedsarebeingenvisaged,inwhichinformation

hastobeconveyedinatime-criticalfashion.

Forexample,arraysofvideosensorsarealready

usedbyoceanographerstodeterminetheevolu-

tionofsandbarsviaimageprocessingtechniques

[58].

•Personlocatorservices.Multimediacontentsuch

asvideostreamsandstillimages,alongwith

advancedsignalprocessingtechniques,canbe

usedtolocatemissingpersons,oridentifycrimi-

nalsorterrorists.

•Industrialprocesscontrol.Multimediacontent

suchasimaging,temperature,orpressure

amongstothers,maybeusedfortime-critical

industrialprocesscontrol.Machinevisionisthe

applicationofcomputervisiontechniquesto

industryandmanufacturing,whereinformation

canbeextractedandanalyzedbyWMSNsto

supportamanufacturingprocesssuchasthose

usedinsemiconductorchips,automobiles,food

orpharmaceuticalproducts.Forexample,in

qualitycontrolofmanufacturingprocesses,

detailsorfinalproductsareautomatically

inspectedtofinddefects.Inaddition,machine

visionsystemscandetectthepositionandorien-

tationofpartsoftheproducttobepickedupby

aroboticarm.Theintegrationofmachinevision

922I.F.Akyildizetal./ComputerNetworks51（2007）921–960systemswithWMSNscansimplifyandadd

flexibilitytosystemsforvisualinspectionsand

automatedactionsthatrequirehigh-speed,

high-magnification,andcontinuousoperation.

Asobservedin[37],WMSNswillstretchthe

horizonoftraditionalmonitoringandsurveillance

systemsby:

•Enlargingtheview.TheFieldofView（FoV）ofa

singlefixedcamera,ortheFieldofRegard（FoR）

ofasinglemovingpan-tilt-zoom（PTZ）camerais

limited.Instead,adistributedsystemofmultiple

camerasandsensorsenablesperceptionofthe

environmentfrommultipledisparateviewpoints,

andhelpsovercomingocclusioneffects.

•Enhancingtheview.Theredundancyintroduced

bymultiple,possiblyheterogeneous,overlapped

sensorscanprovideenhancedunderstanding

andmonitoringoftheenvironment.Overlapped

camerascanprovidedifferentviewsofthesame

areaortarget,whilethejointoperationof

camerasandaudioorinfraredsensorscanhelp

disambiguateclutteredsituations.

•Enablingmulti-resolutionviews.Heterogeneous

mediastreamswithdifferentgranularitycanbe

acquiredfromthesamepointofviewtoprovide

amulti-resolutiondescriptionofthesceneand

multiplelevelsofabstraction.Forexample,static

medium-resolutioncameraviewscanbeenriched

byviewsfromazoomcamerathatprovidesa

high-resolutionviewofaregionofinterest.For

example,suchfeaturecouldbeusedtorecognize

peoplebasedontheirfacialcharacteristics.

Manyoftheaboveapplicationsrequirethesen-

sornetworkparadigmtoberethoughtinviewof

theneedformechanismstodelivermultimediacon-

tentwithacertainlevelofqualityofservice（QoS）.

Sincetheneedtominimizetheenergyconsumption

hasdrivenmostoftheresearchinsensornetworks

sofar,mechanismstoefficientlydeliverapplication

levelQoS,andtomaptheserequirementstonet-

worklayermetricssuchaslatencyandjitter,have

notbeenprimaryconcernsinmainstreamresearch

onclassicalse