Forest Fire Detection Using Artificial Neural Network Algorithm Implemented in Wireless Sensor Netwo文档格式.docx
- 文档编号:5129766
- 上传时间:2023-05-04
- 格式:DOCX
- 页数:12
- 大小:22.81KB
Forest Fire Detection Using Artificial Neural Network Algorithm Implemented in Wireless Sensor Netwo文档格式.docx
《Forest Fire Detection Using Artificial Neural Network Algorithm Implemented in Wireless Sensor Netwo文档格式.docx》由会员分享,可在线阅读,更多相关《Forest Fire Detection Using Artificial Neural Network Algorithm Implemented in Wireless Sensor Netwo文档格式.docx(12页珍藏版)》请在冰点文库上搜索。
criteriadetectionisimplementedbytheartificialneuralnetworkalgorithm.Meanwhile,wehavedevelopedaprototypeoftheproposedsystemconsistingofthesolarbattermodule,thefiredetectionmoduleandtheuserinterfacemodule.
Keywords
forestfiredetection;
artificialneuralnetwork;
wirelesssensornetwork
W1Introduction
irelesssensornetworks(WSNs)havebeenthefocusofresearchoverthepastfewyearsbecauseoftheirpotentialinenvironmentalmonitoring,targettracking,andobjectdetection[1].WSNshavealsobeenstudiedinthecontextofdetectingforestfires,whichthreatenforestresourcesandhumanlife.WSNsarenotcostlyandcandetectforestfiresinrealtime,unlikecurrentdetectionmethodsbasedonhumanobservationandunlikespotweatherforecastsorevensatellitemonitoring.WSNscanalsoprovideinformationaboutenvironmentalconditionswithintheforest,whichisusefulforpredictingforestfires[2].Moreover,forestfiredetectionandpredictionisassociatedwithspecificlocationinformationprovidedbyindividualsensornodes.
AlthoughsomepracticalexperimentshavebeenconductedusingWSNstocollectsenseddatafromaforestfire[3]-[5],therearestillsomechallengestousingWSNsforthispurpose.Afiredetectormaysoundanalarmbasedonasimplethreshold,whichgivesrisetofalsealarmseventhoughthesensingunitofthefiredetectormaybehighlysensitive.Falsealarmsoccurfortwomainreasons:
?
Aphotoelectricsmokesensingunitissensitivetowhiteaerosolparticlesfromasmolderingfirebutalsotodust[6].
Environmentalconditionsintheforestoftenseverelydisturbthenormalbehaviorofthesensingunit.Sunlightandartificiallightareprimarysourcesofinterferencewiththeflame?
sensingunit.
Limitedpowersupplytosensornodesmakesitdifficulttodetectfiresoveralongperiodoftime.Thepotentialenergysourcesforsensornodescanbeclassifiedaccordingtowhethertheystoreenergywithinthesensornodes(e.g.,inabattery),distributepowertothesensornodethroughawire,orscavengeavailableambientpower(e.g,usingasolarbatteryonthesensornode).Consideringthevolumeofthesensornode,mannerofdeployment,andforestconditions,thesolarbatteryisoneofthemostpromisingsourcesofenergyfordetectingforestfiresoveralongperiodoftime.However,existingworksonsolarbatteriesforsensornodes,e.g.,[8]-[13],overlooktheproblemofintermittentsunlightintheforest. Inthispaper,weproposeaforestfiredetectionsystemthatincludesanartificialneuralnetworkalgorithmimplementedinaWSN.Overall,themaincontributionsofthispaperareasfollows:
Themulti?
criteriadetectiondependsonmultipleattributesofaforestfireandisintroducedintoWSNstoincreasetheaccuracyofdetectingaforestfire.
Anartificialneuralnetworkalgorithmisusedtofusesensingdatathatcorrespondstomultipleattributesofaforestfireintoanalarmdecision.
WeintroducetheprincipleoftheproposedsystemaswellasaprototypecomprisingTelosBsensornodesandasolarbatterytopowertheWSN.
2SystemDescription
Forthesakeofclarity,weconsideraWSNwithonlyonebasestationandhundredsofsensornodes.BecauseaWSNwithmultiplebasestationscanberegardedasmultipleWSNs(eachcomprisingonebasestationandcorrespondingsensornodes),theproposedsystemcanalsobeimplementediftheWSNhasmultiplebasestations.Therefore,thereare[n]sensornodesintheWSN,eachdenoted[sj],[1≤j≤n].Aforestfire[f]has[l]attributes,eachdenoted[rfi],[1≤i≤l].Attribute[rfi]canbesensedbythesensingunit[ui].A[ui]onan[sj]isdenoted[uji].Theoutputsensingdataof[uji]isdenoted[oji].Forsimplicity,weassumethat[sj]has[l]typesofsensingunitscovering[l]attributesoftheforestfire.Weuseamultilayerback?
propagationartificialneuralnetworktofusesensingdata[oji].Thetotalnumberoflayersintheartificialneuralnetworkisdenoted[m].Theinputvectorofthe[k]thlayer,[1≤k≤m],isdenoted[Ak-1].Theoutputvectorofthe[k]thlayerisdenoted[Ak].Therefore,[A0]and[Am]representtheinputandoutputoftheartificialneuralnetwork,respectively.Thealarmdecisionisdenoted[ad].
3ProposedForestFireDetectionMethod
Inourproposedsystem,detectionismademoreaccuratebyusingmultiplecriteria,whichmeansthe[ad]isbasedonmultiplecriteriaoftheforestfire.Multi?
criteriadetectionisimplementedbytheartificialneuralnetworkalgorithm.Becauseoftheartificialneuralnetwork,theproposedsystemhaslowoverheadandhasself?
learningcapabilities;
thatis,ittrainsitselftobuilduptherelationsbetweensensingdataandcorrect[ad].
3.1MultiCriteriaDetection
Inasystemthatdependsononeattributeofaforestfiretoraisealarms,thereisahighprobabilityoffalsealarmsbecauseofinherentsystemdrawbacksorexternaldisturbances.Toovercomesuchdrawbacksandcounterexternaldisturbances,thesystemmusttakeintoaccountthemultipleattributesofaforestfire.Thisisreferredtoasmulti?
criteriadetection(Definition1).Withmulti?
criteriadetection,multipleattributesofaforestfirearesensedbydifferenttypesofsensingunit.Therefore,asensingunitthathasbeeninterferedwithcannotraiseafalsealarm.Together,multiplesensingunitsconfirmanalarm.Multi?
criteriadetectionincreasestheaccuracyofdetectingaforestfire. Definition1(Multi?
criteriadetection).Multi?
criteriadetectionisrepresentedasafunctionwithmultiplearguments[rf1,rf2,...,rfl],whichrefertotheattributesofforestfire[f],andone[ad],givenby:
[ad=f(rf1,rf2,...,rfl)]
(1)
Theattributes[rf1,rf2,...,rfl]couldbeanycombinationoftheattributesofaforestfire.Thedirectlysensedattributesofaforestfireareflameandheat,whicharesensedbytheflamesensingunitandheatsensingunit.Theflameemitsvisiblelight,buttheforestfirealsoemitsalotofradiation,thespectraldistributionofwhichistheradiationintensitywithrespecttodifferentwavelengthsandisnotuniform.Intheory,theradiationintensityisdeterminedbythetemperatureofthefire.TheradiationintensityfromablackbodywithrespecttothewavelengthandtemperatureisdescribedbyPlanck’sradiationlaw
(2),where[h]isPlanck’sconstant,[c]isthespeedoflight,[λ]isthewavelength,[k]isBoltzmann’sconstant,and[T]isthetemperature[14].
[I=2hc2λ5exp[hc/λkT]-1]
(2)
Therefore,radiationintensitycanbethebasisfordetectingaforestfiregiventhatthetypicaltemperatureofaforestfireis[600°
C-1000°
C][15].Theultravioletsensingunitandinfraredflamesensingunitworkbydetectingradiationintensity.Otherattributesthatcanbeusedtoidentifyaforestfireincludecombustionproducts.Itiswellknownthataforestfiregivesoffburstsofcarbondioxide,carbonmonoxide,watervapor,anddust.
3.2ArtificialNeuralNetworkAlgorithm
Weusethemulti?
layerback?
propagationartificialneuralnetworkformulti?
criteriadetection.AlthoughdatafusioninWSNshasbeencoveredinmuchoftheliterature[16]-[18],thetopichasnotbeenconsideredinthecontextofforestfires.Amulti?
propagationartificialneuralnetworkiswidelyusedtoemulatethenon?
linearrelationshipbetweenitsinputandoutput.However,computationinthiskindofnetworkisnotcomplexbecausethenetworkisacombinationofneuronsdealingwithsimplefunctions.Moreover,multi?
propagationartificialneuralnetworkiscapaleofself?
learning,whichmeansitcantrainitselftobuilduprelationsbetweentheinputsanddesiredtargets.
3.2.1MakinganAlarmDecision
Withoutlossofgenerality,weassumethatthemulti?
propagationartificialneuralnetworkisimplementedon[sj]with[l]typesofsensingunitsthatcover[l]attributesoftheforestfire.Sensingdata[oji]of[uji]on[sj]correspondsto[rfi]oftheforestfire. [A0=[a01,...,a0l]T](3)
Forclarity,letall[oji],[1≤i≤l]compriseacolumnvector[A0](3),where[a0i=oji].Vector[A0]istheinputtothemultiplelayerartificialneuralnetwork.
Specifically,[A0]istheinputtothefirstlayerofthemulti?
layerartificialneuralnetwork.Inthefirstlayer,[A0]ismultipliedbyweightmatrix[W1]withdimension[s1×
l]andbiasvector[B1],including[s1]neuronsinthefirstlayer.Theintermediatecomputationresultofthefirstlayerisdenoted[N1]andisgivenby:
[N1=W1A0+B1](4)
Then,[N1]issenttotransferfunction[F1],whichmaybealinearornonlinear.Thatis,[F1]maybeahard?
limitfunctionorsigmoidfunctiondependingonthespecificproblemitneedstosolve.Ingeneral,transferfunctionsinthemulti?
layerartificialneuralnetworkareeasytocompute.Transferfunction[F1]operatesoneveryelementof[N1].Theresultoftransferfunction[F1],denoted[A1],istheoutputofthefirstlayer:
[A1=F1(N1)=F1(n11)?
F1(n1s1)]
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- Forest Fire Detection Using Artificial Neural Network Algorithm Implemented in Wireless Sensor Netwo
链接地址:https://www.bingdoc.com/p-5129766.html