Now Accepting Apple Pay

Apple Pay is the easiest and most secure way to pay on StudyMoose in Safari.

paper

P a p e r T itle * (u se sty le : p a p e r title )

*Note:Sub-titlesarenotcapturedinXploreandshouldnotbeused line1:1stGivenNameSurname line2:dept.nameoforganization (ofAffiliation) line3:nameoforganization (ofAffiliation) line4:City,Country line5:emailaddressorORCID line1:4thGivenNameSurname line2:dept.nameoforganization (ofAffiliation) line3:nameoforganization (ofAffiliation) line4:City,Country line5:emailaddressorORCID line1:2ndGivenNameSurname line2:dept.nameoforganization (ofAffiliation) line3:nameoforganization (ofAffiliation) line4:City,Country line5:emailaddressorORCID line1:5thGivenNameSurname line2:dept.nameoforganization (ofAffiliation) line3:nameoforganization (ofAffiliation) line4:City,Country line5:emailaddressorORCID line1:3rdGivenNameSurname line2:dept.nameoforganization (ofAffiliation) line3:nameoforganization (ofAffiliation) line4:City,Country line5:emailaddressorORCID line1:6thGivenNameSurname line2:dept.

nameoforganization (ofAffiliation) line3:nameoforganization (ofAffiliation) line4:City,Country line5:emailaddressorORCID Abstract Thisprojecthelpstheblindtoidentifythethings andpeopleinfrontofhimandturnthevisualsintosoundsaswell astoidentifythedistancebetweenhimandthesethingsandpeople inRealtime.Inthisprojecttwoobjectswereusedtoobject recognitionandrecognizepeopleusingfacerecognition.Inobject detectionwasusedYOLOv3modelandInfacerecognitionwas usedlocalbinarypatternshistograms(LBPH)algorithm,along withfacedetectionbyHaarfeature-basedcascadesanddistance- basedclustering. Keywords objectrecognition,facerecognition,YOLOv3, LBPH,Haarfeature-basedcascades I. INTRODUCTION(HEADING1) Visionisoneofthemostimportanthumansensesand playsapivotalroleintheperceptionandinteractionofthe human environment.A visually impaired person requires assistanceforeverydaychores.Itisextremelydifficultfor blind people to navigate in an environmentthatisnot habitual to them[2].According to the World Health Organization(WHO),therearearound285millionvisually impairedintheworld.Amongthe 285millionpeople,39 millionpeoplearecompletelyblindand246millionpeople havelow vision[1].

Get quality help now
writer-Charlotte
Verified writer

Proficient in: Other

4.7 (348)

“ Amazing as always, gave her a week to finish a big assignment and came through way ahead of time. ”

+84 relevant experts are online
Hire writer

WithcurrentadvancementsinComputer visionstate-of-art,wecannow perform imageprocessing anddeterminethecontentsoftheimageoftheimageby classifying objectsinto known classes[3].Through this project,whichwillenableblindimpairedtoidentifyobjects andpeopleinfrontofthem byconvertingthenameofthe objectorperson identified to voice and determine the direction and distance between objects and visually impaired.Inthispaper,wediscussedacoupleoftechniques whichisdesignedtohelpthevisuallyimpairedpeople.The firstone isobjectdetection.Objectdetection isan old fundamentalproblem inimageprocessing,forwhichvarious approacheshavebeenapplied.Butsince2012,deeplearning techniquesmarkedlyoutperformedclassicalones[4].Herewe areusingadeeplearningmethodcalledYOLOv3basedon Darknet53.whichisthelatestversionoftheYou-LookOnly- OnceapproachproposedbyJosephRedmonin2018[4].the secondisfacedetectionandrecognition.Facedetectionis derived from objectdetection using Haarfeature-based cascadeclassifierwhichwasproposedbyPaulViolaand MichealJones[5].Thisisamachinelearningapproach.Face recognitionusedLocalBinaryPatternAlgorithm .LBPH is oneofthemostpopularconventionalmethods;itisusedfor robustdatarepresentation,aswellashistograms,forfeatures reduction[6]. II. METHODOLOGY Thesesmartglassesaredesignedtorecognitionobjects andpeoplearoundtheblindanddeterminethedistanceand directionoftheseobjectsaboutthem andmakeitaslabel. TheprojectcomprisesoftheRaspberry Pi3 which has Raspberry cameraand headset.

Get to Know The Price Estimate For Your Paper
Topic
Number of pages
Email Invalid email

By clicking “Check Writers’ Offers”, you agree to our terms of service and privacy policy. We’ll occasionally send you promo and account related email

"You must agree to out terms of services and privacy policy"
Check writers' offers

You won’t be charged yet!

TheRaspberry camerais usedtocaptureframes.Thisframesispassedtotheobject detectionalgorithm.Deeplearningalgorithmsareourfirst choiceinthistaskinsteadoftraditionalmethods[7].Object Detectioncanbeusedtodetectwithboundingboxesobjects inimages.Intheeventthattheselectedobjectisaperson,a face detection and recognition algorithm will be implementedtodeterminewhopersonis.machinelearning algorithmsareourchoiceinthistask.Thenwedeterminethe directionoftheobjectIsintherightorleftoftheRaspberry cameraanddeterminethedistancebetweenthisobjectand thiscamera.TheRaspberryPiisconverttexttovoiceusing the"pyttsx3"moduleforpython.headsetisusedtohearthe voiceobjectorpersonname,directionanddistanceThat werecalculatedusingthepreviousalgorithms. A. Rasbperrypi3 B. Rasbperrycamera C. Headphone D. powersupply III. OBJECTDETECTIONALGORITHM Objectdetectionisanoldfundamentalproblem inimage processing,forwhichvariousapproacheshavebeenapplied. But since 2012, deep learning techniques markedly outperformed classical ones[4].in this project we used yolov3algorithm forobjectsdetectioninrealtime.itisusing deeplearningtechniques.Thisalgorithm ischaracterizedbyXXX-X-XXXX-XXXX-X/XX/$XX.00©20XXIEEE

speedandaccuracyandacceptstheinputimagesindifferent sizes.YOLOv3 isan objectdetectorproposed by Joseph Redmon, Ali Farhadi[10].YOLOv3 use Darknet-53 for performing feature extraction[10].tuses 53 convolution layersto extractfeaturesfrom theimageand tconsists mainlyof3*3and1*1filters.asillustratedinFigure2. Fig.2.Darknet-53 YOLOv3;Thisalgorithm hasaddedmanyadditionstothe previous version of this algorithm, the second version(yolov2)inordertoimprovetheprocessofobject detection.YOLOv3Thisalgorithm predictsboxesthrough threedifferentscalesandlevelsinsteadofoneasinthe previousversion (YOLOv2),ItusesDarknet-19,butthe additionofresidualconnectionsandfeaturemapmerging makeitlarger,53convolutionallayers,andhenceitiscalled Darknet-53[6].lnCombinedobjectdetectionanddistance estimationthereisOnemethodthatisusingsimilarconcept asoneproposedinthisworkisDisNet[6].Itusesaneural networkwith3hiddenlayersinconjunctionwithYOLOv3 topredictdistanceofadetectedobjectfrom thecamera.A featurevectorismadefrom outputofYOLOv3andisfedto DisNettopredictthedistanceofthedetectedobject[6].This algorithm predictsanobjectivescoreforeachboundingbox usinglogisticregression,butthisshouldbe1ifthereisan overlap from thebounding boxprior thetruthobjecton ground more than another bounding box prior.This algorithm predicts boxes in three differentscales.The system thenextractsfeaturesfrom thesescalesbyusinga conceptvery similarto the pyramid networks feature concept.At320 ? 320 YOLOv3 runsin 22 msat28.2 mAP[5],Thesevalues??areasaccurateasSSD butthese values??are3timesfasterthanSSD.Whenwelookattheold .5IOU mAPdetectionmetricYOLOv3isquitegood.It achieves57.9AP50in51msonaTitanX,comparedto 57.5AP50in198msbyRetinaNet,similarperformancebut 3.8? faster[5].The training data is extracted from the algorithm ofthealgorithm from thereusing custom text software.Thisnetworkhasbeenmodifiedtopredictmore than one operation not only to predict humans and surroundingboxes,butalsotopredictthedistancefrom the cameratothetargetobject. A. WORKINGOF YOLOv3 InitiallytheYOLOv3imageiscapturedandthenthe YOLOalgorithm isappliedsothattheimageisdividedinto networksof3?3multi-matrices.Theimagecanbedivided intoanynumberofnetworks,dependingonthecomplexity oftheimage.Whentheimagedivisioniscomplete,each networkissubjectedtotheclassificationandlocalizationof theobject.Iftheappropriateobjectisnotfoundinthegrid, thevalueoftheobjectandtheboundingboxwillbezero. Thisalgorithm canalsobeusedtopredicttheveryprecise squaressurroundingtheimagebydividingtheimageintoN ?Ngridsbypredictingthesurroundingsquaresofeachgrid intermsofapplyingbothimageclassificationtechniquesand objectlocalizationtoeachgridofimages[2].Thisalgorithm checkseachgridindividually,identifiesthelabelcontaining theobjectandidentifiestheboxessurroundingit.Non-object gridlabelsaremarkedwithzero[2]. Number equations consecutively.Equation numbers, withinparentheses,aretopositionflushright,asin(1),using arighttabstop.Tomakeyourequationsmorecompact,you mayusethesolidus(/),theexpfunction,orappropriate exponents.Italicize Roman symbols for quantities and variables,butnotGreeksymbols.Usealongdashrather thanahyphenforaminussign.Punctuateequationswith commasorperiodswhentheyarepartofasentence,asin: a???b????? ??? Notethattheequationiscenteredusingacentertabstop. Besurethatthesymbolsinyourequationhavebeendefined beforeorimmediatelyfollowingtheequation.Use”(1)”,not “Eq.(1)”or”equation(1)”,exceptatthebeginningofa sentence:”Equation(1)is…” B. BOUNDINGBOXPREDICTIONS YOLO algorithm is used for predicting the accurate boundingboxesfrom theimage.TheimagedividesintoSx Sgridsbypredictingtheboundingboxesforeachgridand classprobabilities.Both imageclassification and object localization techniquesare applied foreach grid ofthe imageandeachgridisassignedwithalabel.Thenthe algorithm checkseachgridseparatelyandmarksthelabel whichhasanobjectinitandalsomarksitsboundingboxes. Thelabelsofthegirdwithoutobjectaremarkedaszero. IV. FACERECOGNITION Nowadays,facerecognitionhasbecomeoneofthemost importanttopicsincomputervisionthroughmanyimportant applications such as security applications,surveillance applications,applicationsusedinbankingservicesandother applications …. But there are many difficulties and challenges thataccompany this topic is accuracy And efficiency.Itisoneofthemostimportantchallengesthat mustbeachievedtoensurethesuccessoftheseapplications

over the years has been developed many different algorithmstoidentifythefacebymanyresearchersand these algorithmslike Sparse Coding (SC) algorithm , HistogramsofOrientedGradients(HOG)algorithm,Local Binary Pattern (LBP) algorithm, Linear Discriminant Analysis (LDA) algorithm [8],and Gabor feature(GF) algorithm [5].Allthesealgorithmshaveanaccuracyrate between50% to76%[8].Butthereisanalgorithm that identifiesthefrontfaceandthelateralfacealsowitha90% accuracy.system includes four main parts: information acquisitionmodule,featureextractionmodule,classification moduleandtrainingclassifierdatabasemodule[8].Herethe picture information collected through the Learning AcquisitionUnitwasusedasatestsampleforanalysis.In thefeatureextraction unit,which representsthehuman identity information is extracted and examined. The classificationunitusestheclassification,whichwastrained byadatabasetoclassifythetestsamplesinordertoidentify allpersonidentificationinformation. A. FACEDETECTION We have used OpenCV which presentsa Haarcascade classifier[8],[8],whichisusedforfacedetection.TheHaar cascadeclassifierusestheAdaBoostalgorithm to detect multiplefacialfeatures.First,itreadstheimageto be detectedandconvertsitintothegrayimage,thenloadsHaar cascadeclassifiertodecidewhetheritcontainsahumanface. Ifso,itproceedstoexaminethefacefeaturesanddraw a rectangular frame on the detected face.Otherwise,it continuestotestthenextpicture[8]. B. FEATUREEXTRACTION Herethecontrastinformationofthepixelstotheiradjacent pixelsisdescribed by applying the LBP operator.The originalLBPoperatorisdefinedina3*3window.Asa maximum window,theaveragepixelvalueisused,andis thencomparedtothegrayvalueoftheadjacent8-pixelpixel. Withoutthisitismarkedwithavalueof(0).Thefunctionis definedasshowninequation1[8].Itcanbeillustratedin Figure2 Byusingthismethod,eight3*3pointsarecomparedto create 8-bitbinary numbers.When these numbers are changedtodecimalnumbers,theLBPalgorithm values??are obtainedforthelazinesspointsinthemiddleofthewindow, whichareusedforrenderingregiontexturefeatures,andin theLBPH algorithm animprovedLBPcircularoperatoris used.Text heads organize the topics on a relational, hierarchicalbasis.Forexample,thepapertitleistheprimary text head because all subsequent material relates and elaboratesonthisonetopic.Iftherearetwoormoresub- topics,the nextlevelhead (uppercase Roman numerals) shouldbeusedand,conversely,iftherearenotatleasttwo sub-topics,thennosubheadsshouldbeintroduced.Styles named “Heading 1”, “Heading 2”, “Heading 3”, and “Heading4″areprescribed. C. FiguresandTables a) PositioningFiguresandTables:Placefiguresand tablesatthetopandbottom ofcolumns.Avoidplacingthem inthemiddleofcolumns.Largefiguresandtablesmayspan acrossbothcolumns.Figurecaptionsshouldbebelow the figures;tableheadsshouldappearabovethetables.Insert figuresandtablesaftertheyarecitedinthetext.Usethe abbreviation”Fig.1”,evenatthebeginningofasentence. TABLEI. TABLETYPESTYLES Table Head TableColumnHead Tablecolumnsubhead Subhead Subhead copy Moretablecopya a. SampleofaTablefootnote.(Tablefootnote) Fig.1.Exampleofafigurecaption.(figurecaption) FigureLabels:Use8pointTimesNewRomanforFigure labels.Usewordsratherthansymbolsorabbreviationswhen writingFigureaxislabelstoavoidconfusingthereader.As an example, write the quantity “Magnetization”, or “Magnetization,M”,notjust”M”.Ifincludingunitsinthe label,presentthem withinparentheses.Donotlabelaxes only with units.In the example,write “Magnetization (A/m)”or”Magnetization{A[m(1)]}”,notjust”A/m”.Do notlabelaxeswith a ratio ofquantitiesand units.For example,write”Temperature(K)”,not”Temperature/K”. ACKNOWLEDGMENT(Heading5) Thepreferredspellingoftheword”acknowledgment”in Americaiswithoutan”e”afterthe”g”.Avoidthestilted expression”oneofus(R.B.G.)thanks…”.Instead,try”R. B.G.thanks…”.Putsponsor acknowledgments in the unnumberedfootnoteonthefirstpage. REFERENCES [1] "Globaldataonvisualimpairment",WorldHealthOrganization,2019. [Online]. Available: 27-Oct-2019]. [2] S.Jain,S.Varsha,V.Bhatand J.Alamelu,"Design and ImplementationoftheSmartGlovetoAidtheVisuallyImpaired", 2019 InternationalConference on Communication and Signal Processing(ICCSP),2019.Available:10.1109/iccsp.2019.8698009 [Accessed28October2019]. [3] ShitalPatil,A.R.(2019).E-VISIONEyesforVisuallyImpaired usingSmartphoneImplementing.PramanaResearchJournal. [4] AdelAmmar,A.K.(16Oct2019).AerialImagesProcessingforCar Detection using ConvolutionalNeuralNetworks:Comparison betweenFasterR-CNNandYoloV3.arxiv. [5] R.Agarwal,R.Jain,R.RegunathanandC.PavanKumar,"Automatic AttendanceSystemUsingFaceRecognitionTechnique",Proceedings ofthe2ndInternationalConferenceonDataEngineeringand Communication Technology, pp. 525-533, 2018. Available: 10.1007/978-981-13-1610-4_53[Accessed28October2019]. [6] M.AbuzneidandA.Mahmood,"EnhancedHumanFaceRecognition

UsingLBPHDescriptor,Multi-KNN,andBack-PropagationNeural Network",IEEEAccess,vol.6,pp.20641-20651,2018.Available: 10.1109/access.2018.2825310[Accessed28October2019]. [7] S.Luo,C.XuandH.Li,"AnApplicationofObjectDetectionBased on YOLOv3 in Traffic", IEEE, p. 69, 2019. Available: doi.org/10.1145/3317640.3317657[Accessed29October2019]. [8] AftabAhmed,J.G.(2019).LBPHBasedImprovedFaceRecognition AtLowResolution.Chengdu:IEEE. [9] T.Chen,Y.Wotao,S.Z.Xiang,D.Comaniciu,andT.S. Huang,”Totalvariationmodelsforvariablelightingfacerecognition” IEEETransactionsonPatternAnalysisandMachineIntelligence, 28(9):1519{1524,2006. IEEEconferencetemplatescontainguidancetextfor composingandformattingconferencepapers.Please ensurethatalltemplatetextisremovedfrom your conferencepaperpriortosubmissiontotheconference. Failuretoremovetemplatetextfrom yourpapermay resultinyourpapernotbeingpublished. Wesuggestthatyouuseatextboxtoinsertagraphic (whichisideallya300dpiTIFForEPSfile,withallfonts embedded)because,inanMSW document,thismethodis somewhatmorestablethandirectlyinsertingapicture. To have non-visible rules on yourframe,use the MSWord”Format”pull-downmenu,selectTextBox> ColorsandLinestochooseNoFillandNoLine.

Cite this page

paper. (2019, Nov 28). Retrieved from http://studymoose.com/paper-801-example-essay

👋 Hi! I’m your smart assistant Amy!

Don’t know where to start? Type your requirements and I’ll connect you to an academic expert within 3 minutes.

get help with your assignment