基于k-均值算法的无监督模型降雨预报.(IJMSC-V1-N1-2)

1、I.J.MathematicalSciencesandComputing,2015,1,11-20PublishedOnlineJuly2015inMECS()DOI:10.5815/ijmsc.2015.01.02Availableonlineat:Rainfallprediction,Gaussianmixturemodel,K-Meansalgorithm,rainfallestimation,PSNR,MSE.©2015PublishedbyMECSPublisher.Selectionand/orpeerreviewunderresponsibilityoftheResearchAssociationofModernEducationandComputerScience1.IntroductionWeatherforecastingisconsideredtobeoneofthemostrecurrentchallengingissues.Theprosperityandeconomicscenarioofacountrycompletelydependsontherain。

2、fall.Duetotheincreaseinglobalheats,itisinevitabletopredicttherainfallaccurately.Thepredictionofrainfallisgenerallybasedonthecloudformulation.Thecloudsareclassifiedintothreegroups,namely;nimbostratus,cumulonimbusandcumulus.Amongtheseclouds,thefirsttwotypesofcloudsproducerainfallandtherainfallassociatedwiththethirdtypeofcloudisrare.Thereforepredictivemethodologiesaretobeimplementedfortheidentificationofcloudandtherebyestimatingtheprobabilityofrainfall.Manymodelshavebeendiscussedintheliteratureabou。

3、tthecloudformationandpredictionoftherainfall.Ingeneral,theonsitepredictionofrainfalliscarriedoutintwostages,stage-1discussesaboutthedatasetandextractionofthefeaturesfromthedataset.Inthestage-2,detectionofthecloudisestimated.Inordertoportraythepresentmethodology,thedatasetisconsideredfrom*Correspondingauthor.Tel.:E-mailaddress:vamsikrishna527@gmail.com12PredictionofRainfallUsingUnsupervisedModelbasedApproachUsingK-MeansAlgorithmsatelliteimagesfromthemeteorologicaldepartment,Visakhapatnam.Thestudy。

4、isbasedontheinfraredvisibleregionsandwatervaporregions.TheseimagesareextractedfromINSAT-3AandKALPANA-1.Fortheclouddetection,thesatelliteimagesarepreprocessedandbasingonthehighreflectanceofthecloudinthevisiblespectrumandlowtemperaturesininfraredspectrum,theanalysisiscarriedout.Fromthesatelliteimages,thecloudtoptemperatureisextractedusingtheinfraredraysandbasingontheprecipitation,onecananalyze,whetherthecloudisacold-cloudorahot-cloud.Intheabsenceofcloud,thesatelliteimagesprocessthedataandrecordthe。

5、dataasinfraredintensity.Clouddensityhelpstounderstandthethicknessofthecloudwhichindirectlydecidesintheformationoftherain.ItisgenerallyconsideredasaBooleanwhichreturnsazero,whencloudisnotpresentandintensityvaluewhichisequaltoone,ifcloudispresent.Watervaporisanothersignificantparameterthatdecidesthestatusoftherainfall.ThearchitectureinvolvedispresentedbelowinFig.1Fig.1.ArchitectureTheresearchworkinthisareahasbeencarriedoutbymanymeteorologistsandmostofthemodelsdevelopedbytheeminentresearcherscouldn。

6、otabletopredicttherainfallefficiently,becauseofthecompressionratiowhichisassociatedwiththeretrievedimages.Theotherfactorsassociatedwiththesatelliteimagesinclude;snow,sandandicyregions,theseassociatedfactorsaffecttheretrievalaccuracy.Thereforeeffectivemethodologiesarestilltobeinvolvedforeffectiveidentificationofcloudandformationofcloud.Therefore,thispaperhighlightsamodelbasedonfeatureextractiontogetherwithclusteringtechniquesVisibleImageInfraredImageWaterVaporImageCloudDensityCloudTopTemperatureS。

7、eaSurfaceTemperatureHumidityEstimationFusedAND/ORRainfallPredictionDatabasePredictionofRainfallUsingUnsupervisedModelbasedApproachUsingK-MeansAlgorithm13basedonk-meansalgorithm.Theclassificationofcloudsisbasedontheparameterswhichincludewatervapor,humidityandtheotherparametersestimatedfromGaussianmixturemodel.Therestofthepaperisorganizedasfollowssection-2ofthepaperdealswithrelatedworkinthearea;section-3presentsthefeatureextractionmethodology.TheclusteringtechniqueforsegmentingtheimagesbasedonK-Me。

8、ansalgorithmispresentedinsection-4,inthesection-5ofthepaperhighlightsabouttheclassificationalgorithmbasedonGaussianmixturemodel,section-6ofthepaperdiscussesaboutthefusionandthepredictionoftherainfalliscarriedoutbyusingKL-divergencewhichishighlightsinsection-7ofthepaper.Insection-8thedesignandimplementationisshown.Insection-9evaluationishighlightedandsection-10ofthepapershowsexperimentsandresultsandthenthepaperconcludes.2.RelatedWorkManyauthorshavepresentedthereviewofthepredictionofrainfall.D.K.R。

9、ichardsandG.D.Sullivanhaspresentedamethodologyinwhichthepredictionoftherainfallisbasedonthediscriminationofthecloudforwhichtheclassifi。