Real Time & Automated Meter Reading using Image Processing Reducing Human Error

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN:2395-0072

Real Time & Automated Meter Reading using Image Processing Reducing Human Error

1,2,3,4Student, Dept. of Electronics Engineering, Vivekanand Education Society’s Institute of Technology, Mumbai, Maharashtra, India

5Associate Professor, Dept. of Electronics Engineering, Vivekanand Education Society’s Institute of Technology, Mumbai, Maharashtra, India ***

Abstract - Utilities (Gas, and Electricity) plays a pivotal role in our everyday life. To keep track of energy consumption meters are used. Earlier Analog meterswereusedwhich posed some problems like the parallax error, the needles getting stuck, drifting out of calibration, susceptible to shock and vibrations. Eventually, with the replacement of analog meters with digital meters, these problems were solved. However, a common disadvantage of analogue meters and digital meters was the need for workers to be physically present to record readings from the meter and subsequently relay those values to the utility provider firm. This conventional method is susceptible to errors and the users have to pay more than what they actually had to pay.

The focus is to solve this problem with a system that automatically captures the image of the meter and sends it to the server wherein image processing is done to obtain the meter readings. Another advantage of the system is the reduction of human resources and expenses for recording meter readings physically.

The setup will include a webcam that will capture images at predetermined intervals. Raspberry Pi Zero W is used to send these clicked images to the Firebase Realtime Database. This image is then retrieved to extract the data regarding the utility consumption by using a Python library called EasyOCR. To increase the efficiency of extracting the correct data from the snapshot of an energy meter, softwaretoolssuchasGoogle Vision API, Amazon Rekognition, and Node-RED are used.

theirworkforce,checksthenumbersshownintheelectric meter panel periodically, usually every month. This procedure that involves noting down the readings of the meter,isdonemanually.Theoperatorcapturesanimageof the energy meter and the readings for the same are recorded. The operator reports the findings back to the utility provider and the provider will calculate how much eachcustomerwillbecharged.Theentireproceduretakesa longtimeandisvulnerabletohumanerror.Moreover,itis not feasible to perform this manual process for the utility meterssetupinruralareasDuetotheincreasedpopulation and industrialization in countries such as India, a better method for real time and expedient meter reading is proposedinthispaper.

1.1 LITERATURE REVIEW

Key

Words: Analog Meters, Raspberry Pi Zero W, EasyOCR, Node-RED, Amazon Rekognition

1. INTRODUCTION

Utilities(Gas,andElectricity)areessentialservices that play a major role in our lives. With the rise in productivityandmodernization,theusageoftheseutilities is increasing every day. In the case of electricity billings, consumersoftenencounterissueswithdefectivebills,which thenbecometheirsoleproblem,requiringthemtolookout for a resolution from the MSEB. Human intervention in collectingthereadingsofthemeter,istheonlyexplanation forthat issuesince,theutilityprovidercompany,through

In previous works, many models were designed using differentimageprocessingtechniquesandalgorithms.In[1] the author proposes a system that takes automatic meter readingswiththeaidofa sensor.Acamera ismounted in frontofthehouse'senergymetertocapturephotographs. Contouralgorithmisusedwithaprocessortoseparatedigits andmeasurethebillforthemonth.TheRaspberryPiisused inthispaperbecauseitisakindofminicomputer.Itisnot possible to run Microsoft Windows on it because it has a differentprocessor,henceitisrecommendedtouseLinux operating systems that are very similar to Windows. RaspberryPiisusedtosurftheinternetandsendanemail abouttheelectricityconsumptionofthecurrentmonth.The billisthenwirelesslytransmittedtotheserverviaGSMand shownonanLCDfortheuser'sreference.Accordingtothe conceptofAMR,itallowsforeasysavingsbymeterreading, increaseddataaccuracy,fasterbilling,andbettercustomer service. The camera is placed in front of the E-meter. The camera takes a picture when the order is sent. The bill is calculated and sent using GSM using this image, which is processedbytheRaspberryPiusingthecontouralgorithm.

Another attempt [2] is made in this work, which offers a systemthatemploysmobilephonestotakeapictureofthe power meter. Image processing functions are then implementedtoautomaticallyidentifythemeterreading's digits. Preprocessing, Digit Segmentation, and Digit

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN:2395-0072

Recognitionarethethreekeyprocessesthatthisimagegoes through. In the Preprocessing Stage, theRGB image is convertedtoagrayscaleimage,whichisfollowedbyImage Binarization.Anoisereductionalgorithmisusedtonegate thenoiseproducedintheimage.Later,themeterreading’s area in the binarized image is cropped. Next comes the second stage of Data Segmentation that aims to create six segmenteddigitsbyscanningtheclippednumericregionleft to right, vertically and horizontally. In order to make the recognition stage easier, all segmented digits are finally resizedtomakethemallroughlythesamesize.Nextisthe stageofDigitRecognitioninwhichonesegmentedsampleof eachdigitisusedtoextractthefeaturesofthatdigitandis storedasatemplate.Henceeverydigitwillhaveitsunique template.Thesetemplatesarethenusedtoidentifythedigits fromthesegmentedimagesbymatchingthefeatures.This systemrecognizesthedigitsoftheelectricmeterreadingin threemajorstagesofimageprocessingwithasufficientlevel ofaccuracyof96.49%foreachdigitandanoverallreading accuracyof85.71%.

The authors in [3] propose a system wherein a timer is associatedwithawebcam.Thetimerinstructsthecamerato clicktheimageofthemeteratregularintervalsoftime.The capturedimageisthenconvertedtoabinaryimage,followed byadjustmentsbyalteringthecontrastandbrightness.The meterreadingareaiscroppedbyadjustingthetrimheight andwidthtoobtainthenumericarea.TheSupportVector Machinelearningalgorithmisappliedtothepre-processed imagetodetectandsegmentthedigitsofthemeterreading. TheSupportVectorMachineisthenreappliedtoeachofthe segmentedimagesinordertodistinguishdigitsfrom0to9. Finally, the output and other information, including the nameand numberof the customer,thedateand time, are alsosenttotheserver.Whenacertainamountoftimehas passedwithouttheserverreceivingthemeterreading,the server thinks that the camera is broken and dispatches servicepersonneltoreplacethedamagedcamera.

An Automatic Meter Reading (AMR) system prototype is described in the research work in [4]. Three components make up the prototype design: an interface circuit, a traditionalorelectromechanicalmeter,andZigBeemodules. The design of a cost-effective interface circuit and the programming of ZigBee modules are the most important tasks.ThemodulesaredesignedtobehavelikeEnd-Devices, Routers,andCoordinatorsintheZigBeewirelessnetwork and Concentrators and Range Extenders. According to its characteristics,themeterreadingsystemhasadistributed structure. A data collector, wireless communication networks, and a server system make up the system. Data from digital meters is sent through two boards: a data collectionmoduleandaZigBeetransmissionmodule.Thisis wheretheZigBeetrueSystem-on-ChipCC2430makesabig difference. It combines the excellent performance of the market-leadingCC2420RFtransceiverwitha32/64/128KB

flash memory, 8 KB RAM, an industry-standard upgraded 8051MCU,andaplethoraofadditionalpracticalfeatures.It producesverylittlenoiseandconsumesverylittlefuel.The cameratakesapictureofthemeterreadingandtransmitsit overZig-BeetotheserverPC.There,theimageisidentified andsegmented,andthedigitsarereadtobeusedincreating bills.Themethodusedinthisstudyprocessedimagesusing thefollowingsteps:Afterconvertingtheimagetograyscale, thenumericareaiscropped,projected,andthenbinarized using a threshold. Segmentation and recognition are then carriedoutusingadigitalrecognitionalgorithm.

In [5] the exploration work is tied in with perusing the upsidesoftheelectricalmeterbyutilizingawebcamerathat snaps a picture of the meter, perceives the digits, and afterward stores the yield in a content document. A web cameraof325x288isconnectedtoalaptop.Infrontofthe meter, the camera is mounted. The camera will take an imageoftheelectricitymeterinasequenceoftimeintervals. Then,itwillbetransformedusingthegray-scaleconversion methodintoagrayscaleimage.Thefundamentalideabehind thisconversionistopreservebrightnesswhileremovingthe colorandsaturationinformation.Theidealwaytochange RGBtograyscaleisshowninthefollowingequation.

Luminance=0.299*Red+0.587*Green+0.114*Blue

Thepreviousstage'sgray-scaleimageistransformedintoa binaryimage(Black&White).Thisconversionisthemost crucial step in all phases of the EMRR method, and more specifically in the plate extraction process to resolve illuminationissues.Inthenextstage,theimageistrimmedin such a way that only the meter reading region will be extractedtoidentifythedigitsinthenextstage. Then,use the Vertical Edge Detection Algorithm (VEDA) to segment theimageandrecordtheoutcomeinanarraymatrixafter searchingtheimagehorizontallytillyoulocateawhitepixel. Finally,compareeachareaoftheimagetothetemplates,and thenrecordtheresultsasthemeterreadinginatextfile.

In[6]asystemisproposedwhereinatspecificintervalsthe image is captured by a video camera. After imaging, the softwarecansavetheimagetothefilesystem,whereitcan be read later. Since the camera does not have to be connectedallofthetime,thisisveryusefulforproduction andtesting.Afterthat,thepictureispre-processed.Thisstep should adjust and refine the image so that the individual digitsofthecountercanbedetectedandisolatedinthenext step. The extracted digit images are then fed into the character recognition system (OCR). It must, however, be interactively trained with a selection of all possible characters (the digits 0 through 9). This results in a file containingacollectionoftrainingdata.TheOCRloadsthis trainingdataduringnormalactivityandcanthusidentifyan unknowncharacter.Thisclassificationcomeswithadegree oferror.Itmakessensetocheckthereadingsforplausibility

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN:2395-0072

before proceeding with the processing. If it passes, the observedcountervalueissavedinadatabasealongwiththe current time. Later evaluations, such as the generation of graphicswithhourly,regular,andweeklypowerusage,can bemadeusingthedatastoredhere.

1.1 PROPOSED SYSTEM

TheproposedsystemconsistsofaWebcamwhichisplaced in the sight of the utility meter, as shown in Fig-1. The Webcam is connected to Raspberry Pi Zero W which is powered by a 5 V power bank. At specific intervals, the webcam captures the image of the meter which is subsequentlysenttoFirebaseRealtimeDatabasewiththe help of Raspberry Pi Zero W. The image is then retrieved from the Database and image processing methods are applied.

monitorusingmini-HDMItoHDMIcable. ConnecttheOTG cabletomicro-USBport.Connectmouse,keyboardtoOTG cable. Power the Raspberry Pi Zero W. Connect it to WiFi.Openthe‘Terminal’.Runupdateandupgradecommands to download the package lists from the repositories and update to get information on the newest versions of packagesandtheirdependencies.

Step2: Taking image at regular intervals with automated pythoncode.

Step 3:Pushingtheimagetoserver.ConnectUSBWebcamto Raspberry Pi Zero W and run ‘fswebcam image_name.jpg’ tocapturetheimage.Thenexecutethepythoncodetopush the clicked image to Firebasedatabase so that it can be retrievedforfurtherprocessing

Step 4: Retrieving image from server

The image from firebase is retrieved using python code implementingthepyrebaselibrary. Pyrebaseisaninterface between Python and Firebase’s REST API. It allows us to store and retrieve data, images, etc from firebase. Before retrieving the image, we have to establish a connection between our python code and Firebase which is done by includingtheconfigurationdictionaryorJSON.Weinitialize thisconfigurationusingthefunctioninitialize_app().Images inFirebasearestoredinaspecificsectioncalledStorage.So, toaccessitwecallthestorage()functionfollowedbyadot operator variable that is initialized with the initialize_app function.

Thisstoragefunctionisfurtherusedtoretrievetheimage usingthechildfunctionattachedtoitusingadotoperator withthespecificnameoftheimagestoredinFirebaseasa parameter. Finally, the image is downloaded by using the download function having attributes filename initialized with the name by which it has to be stored and another attributeaspathinitializedwithpathname.

Step 5: Applyingimageprocessingattheserverend

UsingEasyOCR

Fig -2-: Raspberry Pi Setup

Step 1: Setting up Raspberry Pi Zero W with integrated camera module as shown in Fig 2. To begin with, install ‘Raspberry Pi Imager’ software on the Laptop/PC. Take a 32/64GBmicroSDcardandusinganadapter/cardreader, connect it to a Laptop/PC. Run Raspberry Pi Imager and formatthemicroSDcard(makesurethatthemicroSDcardis selected using the ‘Choose Storage’ option). From the ‘Choose OS’ option select ‘Raspberry Pi OS with desktop’. Select the ‘Write’ option. Insert the microSD card into Raspberry Pi Zero W. Connect Raspberry Pi Zero W to a

Fig -3:WorkFlowofthesystemusingEasyOCR

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN:2395-0072

The image from firebase is retrieved using the pyrebase libraryandisstoredinthelocalstorageofoursystem.The necessary part of the image is the portion containing the digits.Sothatjobisaccomplishedbyauser-definedfunction called image_crop().Theimagecurrentlyresidesinthelocal storageofthesystem.Tousethatstoredimageweareusing acomprehensiveimageprocessinglibrarycalled OpenCV Opencvhasanimagereadingfunctioncalledimread()which isusedtoprocesstheimageasshowninFigure4.2.2.The argumentsprovidedtotheimread()functionisthepathto theimageinthelocalsystem imread() functionreadsthe image in the form of a matrix. The next part involves cropping the necessary portion of the image which is manuallydoneconsideringthedimensionsrequired.After the necessary portion is extracted, the further process involvessegregationofeachdigitintheimagewhichisdone by calling another user-defined function called digit_extraction(). The digit extraction function extracts each digit and stores it in the local storage. Each digit is recognizedbyEasyOCRusingeasy.Reader([‘en’])function whereensignifiestheEnglishlanguage.Thereaderfunction isusedincollaborationwiththereadtext()functionwhere imageeachindividualimageissentasanargumentwithan additional argument called allowlist which restricts the recognition between 0 to 9. The recognized digits are displayedtotheuser.

Fig -4:WorkFlowofthesystemusingAmazonRekognition

TheimagewhichisstoredinFirebaseisretrievedusingthe pyrebaselibrarywhichisadedicatedpythonlibrarytostore andretrieveimagesfromFirebase.Theretrievedimageis storedin thelocal storage ofthe system. Afterstoringthe image, wehave to get the AWS credentials to use the text detectionmoduleofAmazonRekognition.Weacquirethese credentialsinCSVformatfromtheIdentityandManagement AccessSection ofAmazon Web Serviceswhere weopt for AmazonRekognition.WereadtheCSVfileinpythontoget the aws_access_key_id and aws_access_secret_key to authenticateAmazonRekognition.AmazonRekognitionAPI isaccessedinpythonusingtheboto3library.Boto3isthe Python SDK for AWS. Boto3 allows us to access AWS

resources from python scripts. It can create, update and deleteAWSresources.Weusetheclientfunctionalongwith the boto3 library to authenticate rekognition. The image storedinthelocalsystemisreadinbinaryformatwhichis providedasadictionaryargumenttodetecttextfunctionof Amazon Rekognition. The response is received in a JSON format. The digits are retrieved iterating through the TextDetections key in the JSON. We can customize the iterationaspertheJSONreceivedandthenumberofdigits toberecognized.Figure4givesusabriefideaoftheprocess explainedabove.

UsingGoogleVisionAPI

Fig -5:WorkFlowofthesystemusingGoogleVisionAPI

The image which is stored in Firebase is retrieved using pyrebaselibrarywhichisadedicatedpythonlibrarytostore andretrieveimagesfromFirebase.Theretrievedimageis storedinthelocalstorageofthesystem.Toauthenticatethe Google Vision API, we are using the API key which is downloaded from Google Developers API. The path of the storedimagehastobedefinedbeforesendingarequestto GoogleVisionAPI. Thispath isprovidedtoa user-defined function called image_request. image_request() function createsaJSONfilewithcontentkeyhavingbase64encoding value.Base64isagroupofbinariestotextencodingschemes thatrepresentbinarydatainASCIIstringformatThebase64 encodingisimplementedbycallingauser-definedfunction encode_image.encode_image()functionreadstheimagefile in binary format and returns base64 encoding using the base64 library used in collaboration with the b64encode function. The JSON file, URL of Google Vision API ("https://vision.googleapis.com/v1/images:annotate"),and API key are given as arguments to the request function whichisaninbuiltfunctionoftherequestlibraryinpython. APOSTrequestisrenderedtotheURLandtheresponseisin JSON format. The digits recognized are then fetched by iteratingthroughtheJSONreceived.Figure5istheworkflow oftheabove-explainedevents

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN:2395-0072

1.2 PROPOSED SYSTEM

A programming tool called Node-RED can be used to establish novel and intriguing links between physical components,APIs,andweb services.Theseveral nodesin thepalettecanbeeasilyconnectedusingthefloweditorthat comes with Node-RED. The flow can then be quickly and simplydeliveredtotheruntime. JavaScriptfunctionscanbe createdinsideofarichtexteditor.Youcansaveandreuse helpful features, models, and flows thanks to a built-in library.

Open ‘Terminal’ and run the ‘node’-red-start’ command. OpenanybrowserandentertheIPaddressofRaspberryPi followedbytheportaddressofNode-REDwhichis‘1880’, separatedbyacolon(:).Itcanbeexpressedas‘http://<ipaddress>:1880’. Install node-red-contrib-usbcamera and node-red-contrib-image-simple-nodefromManagePalettein Node-RED’swebinterface.Draganddropthenodesfromthe node window which is on the left side of the screen and createaflowasshowninFig3.

Trigger (Inject Node) is used for manual/automatic triggeringoftheprocess.Atimercanbeusedtotriggerthe processincertaintimeintervals.Thecameranodecaptures theimageusinga USB Webcam.Itstoresthe imagein the local storage of the Raspberry Pi where the path is to be mentionedintheconfigmenuofthenode.Imagenodecrops theimageaccordingtothecoordinatesdefinedwhichresults togivethereadingregionofthemeter.Thiscroppedimage is also stored in local storage. The file node loads the cropped image and anHTTPrequest issent totheGoogle Vision API to give the readings using the function node. GooglevisionAPIkeyissetinthechangenode.Todisplay themeterreading,afunctionnodeisused.Thefinalmeter readingisdisplayedinthedebugsidebarwhichispresenton therightsideofthescreen.Todisplaymessagesinthedebug sidebar,Debugnodes(Greencolorednodes)areused.

2. EXPERIMENTAL RESULTS

Fig -6-: Node-Red workflow

Fig8showsthecroppedimageoftheGasmeterrecognized byEasyOCR. EasyOCRsuccessfullyrecognizesallthedigits displayed in the adjacent side of the Fig 8. Fig 9 is the experimental result on a blurred image. EasyOCR fails to recognizeallthedigitssuccessfullyforblurredimagewhich isoneofthelimitationsofEasyOCR.Therecognizeddigits aredisplayedontherightinFig9

Fig -7-: SystemBlockDiagram2.0

The Fig 6 shows the workflow of NodeRed. NodeRed is a contingenttotheproposed systemasshowninFig 1. The useofNodeRedeliminatestheneedofserver.Itdoesimage processingatrealtimewhenthetriggerisinitiatedasshown inFig7

Fig -9: Image Processing using EasyOCR for blurred meter image

Input meter image as shown in Fig 10 is rendered as a request to Amazon Rekognition API, The API successfully recognizesallthedigitsasshowninFigure10.Asfarasthe Gasmeterimageisconcernedasshownontherightsideof Fig11whenrenderedasarequesttoAmazonRekognition

International Research Journal of Engineering

and

Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN:2395-0072

APIfailstorecognizealldigits.Itissuccessfulinrecognizing onlythefirstsixdigitsoutof8digitsasshowninFig11.

Node-REDwithRaspberryPiintegrationwithGoogleVision APIasanOCRfordigitrecognitionwassuccessfullyexecuted for real-time meter reading. Fig 13 is the image captured whenwetriggertheprocessflowofNode-Redasshownin Fig6.Fig13showsthecroppedportionofthereadingpart and the output reading that is displayed in the debug sidebar. Node-Red with Raspberry Pi using Google Vision APIgivesrealtimeandexpedientresultswithsatisfactory accuracy.Italsoreducesthedelayandsavesstoragespace requiredforstoringimagesattheserverend.

GoogleVisionAPIissuccessfulinrecognizingawiderangeof utilitymetersoutperformingotherOCRsdiscussedabove. TheGasmeterimageisshowninFig10isgivenasaJSON request to the Google Vision API which sends a JSON responsethatgetsallthedigitsaccurately.Fig10showsthe digitsrecognizedbyGoogleVisionAPIfortheGasmeterand meterimagerespectively.TheperformanceofGoogleVision APIforawiderangeofmeterswasbetterthanEasyOCRand AmazonRekognition.Itwassuccessfulingivingpreciseand convincingresults.

3. CONCLUSIONS

Utilitymetersplayasubstantialroleinourlives.Accurate meter readings are of paramount importance. Manual submission of these readings may cause errors. To avoid such errors our proposed system is implemented. Our system prevents any human intervention. We were successfulinachievingthedesiredgoalusingRaspberryPi ZeroWintegratedwithWebcaminassociationwithNodeRed incorporating Google Vision API. Google Vision API stands out as the most efficient OCR recognition with impeccableaccuracyforvariedutilitymeters

4. FUTURE SCOPE

Thefuturescopeofthepaperincludesenhancingsecurity compliances. Creating an application which integrates the hardware implementation and software implementation displayingthedataatrealtimeistheprimaryfuturisticgoal of this project. Enhancing Data Visualization models for better contemplation of data. Advanced Deep Learning modelstopredictfuturisticdatatopreventmalpractices.

ACKNOWLEDGEMENT

Wewouldliketothankourprojectguide Prof. Dr. Abhay Kshirsagar.Thisprojectworkwouldnothavebeenpossible without the encouragement and the able guidance of our guide.Hisenthusiasmandoptimismmadethisexperience bothrewardingandenjoyable.Wewouldalsoliketothank

International Research Journal of Engineering and Technology (IRJET) e-ISSN:2395-0056

Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN:2395-0072

ourHeadoftheDepartmentofElectronicsDepartment,Prof. (Mrs.)KavitaTewari.Iwouldalsoliketoacknowledgethe guidance and support of our Project Coordinator Mr. Anurag Chugh.Hisfeedbackandeditorialcommentswere alsoinvaluable.Wewouldliketoextendsincerethanksto Principal Dr. (Mrs.) J.M. Nair, V.E.S.I.T. for her invaluable support. We would like to express our sincere thanks towards the entire staff of the Electronics Department for theircooperationinthecompletionofourproject.

REFERENCES

[1] SnehaSalunkhe,Dr.(Mrs.)S.S.Lokhande,"RaspberryPi basedautomaticmeterreading",InternationalJournal of Application or Innovation in Engineering & Management (IJAIEM),Volume 5, Issue 9, September 2016,pp.069-073,ISSN2319-4847.

[2] A.Elrefaei,A.Bajaber,S.Natheir,N.AbuSanab,andM. Bazi, "Automatic electricity meterreading based on image processing," 2015 IEEE Jordan Conference on Applied Electrical Engineering and Computing Technologies (AEECT), Amman, Jordan, 2015, pp. 1-5, DOI:10.1109/AEECT.2015.7360571.

[3] C. Edward, Support Vector Machine Based Automatic Electric Meter Reading System, IEEEInternational Conference on Computational Intelligence and ComputingResearch,2013,pp.1-5.

[4] M. Shinde and P. Kulkarni, Reading of Energy Meter basedonImageProcessingTechnology,

InternationalJournalofElectronicsCommunicationand Computer Engineering, Volume 5,Issue (4) July, Technovision-2014,pp.1-5.

[5] K. Parthiban and A. M. Palanisamy, Reading Values in Electrical Meter Using ImageProcessing Techniques, IEEEInternationalConferenceonIntelligentInteractive SystemsandAssistiveTechnologies,Coimbatore,INDIA, 2013,pp.1-6.

[6] Elias Farah, Isam Shahrour, “Smart Water Leakage Detection:FeedbackAbouttheuseofAutomatedMeter Reading Technology”, Sensors Networks Smart and EmergingTechnologies2017.

[7] Champ Prapasawad, Kittitachpornprasitpol ,Wanchalermpora,“DevelopmentofanAutomaticmeter reading system based on ZigBee pro smart energy profile IEEE 802.15.4 standard”, International ConferenceonElectronicDevicesandSolid-StateCircuit (EDSSC),pp.1-3,Dec2012.

[8] NajmusSaqibmalik,Friedrichkupzog,MichaelSonntag, “An approach to secure mobile agents in automatic

meter reading”, IEEE, International Conference on Cyberworlds,computersociety,pp.187-193,2010.

[9] Abhinandan Jain, Dilip Kumar, JyotiKedia, “Smart and intelligentGSMbasedautomaticmeterreadingsystem”, International Journal of Engineering Research & Technology (IJERT), ISSN: 2278-0181, Vol2, Issue3, pp.1-6,May2012.

[10] H.G. Rodney Tan, C.H. Lee and V.H. mork,” Automatic power meter reading systems using GSM network”. IEEE,8thInternationalPowerEngineeringConference, pp.465-469,2007.