Automatic energy meter reading using image processing

AUTOMATIC ENERGY METER READING

USING IMAGE PROCESSING

SICET DEPT.OF EEE PAGE NO:1

 Automatic energy meter reading using image processing

ACKNOWLEDGMENT

With great pleasure I want to take this opportunity to express our heartfelt gratitude to all the people who
helped in making this Technical seminar a grand success.
First of all I would like to thank Dr. JOSEPH PRABHAKAR WILLIAMS, Head of the Department of
Electrical & Electronics Engineering, for being moral support throughout the period of our study in SICET.
I am highly indebted to Principal Dr. P. MALLESHAM for giving us the permission to carry out this
project. I would like to thank the Teaching & Non- teaching staff of Department of Electrical & Electronics
Engineering for sharing their knowledge with us.
I also express our sincere thanks to Mr. R. VENKAT RAO, Chairman, Sri Indu group of institutions, for
his continuous care towards our achievements.

B KONDAL REDDY
16D41A0210

SICET DEPT.OF EEE PAGE NO:2

 Automatic energy meter reading using image processing

DECLARATION

This is to certify that work reported in the present thesis titled “MODELLING ANALYSIS OF A SOLAR
THERMAL POWER PLANT” done by us in college institution. No part of this is copied from
books/journal/internet and where the portion is taken the same has been duly referred in the text. The reported
work is based on the work done entirely by us and not copied from other source.

Signature of the student

…………………….

SICET DEPT.OF EEE PAGE NO:3

 Automatic energy meter reading using image processing

AUTOMATIC ENERGY METER READING USING IMAGE

PROCESSING

ABSTRACT:
Automation is the concept in which the entire process reduces the time and cost of man work to an immense
limit. Through the introduction of technologies into man force, the results turn up to be in efficient accuracy with
greater speed. The idea of implementing automation in electricity billing process is to show the key features on
efficiency in its results. The proposed system shows that rather than manual procedure of taking electricity meter
reading, we can implement a process in which the meter reading is captured with the help of a image capturing
device. By applying the Edge Detection technology on captured image, we can extract the reading from the image
and it is sent to the electricity board with the help of a network connection. The user can also track their daily
electricity usage with help of an application that can be installed and used. This idea reduces the difficulty of the
employee and increases the efficiency. Thus while implementing it would be at low cost with the desired results
as the consumer wishes.

SICET DEPT.OF EEE PAGE NO:4

 Automatic energy meter reading using image processing

INDEX

CHAPTER 1: INTRODUCTION 7
CHAPTER 2: PROPOSED SYSTEM 8
CHAPTER 3: MODULE DESCRIPTION 10
CHAPTER 4: SYSTEM DESIGN 14
CHAPTER 5: IMAGE PROCESSING 15
CHAPTER 6: RESULTS 18
CHAPTER 7: CONCLUSION 19
CHAPTER 8: REFERENCES 20

SICET DEPT.OF EEE PAGE NO:5

 Automatic energy meter reading using image processing

LIST OF FIGURES

Fig. 1. Training set 09

Fig. 2. Output after Image Processing and Digit Recognition 10
Fig. 3. Image processing module 11
Fig. 4. Arduino module 12
Fig. 5. Website module 12
Fig. 6. Mobile Application module 13
Fig.7. Output that is viewed by consumer. 13
Fig. 8. System design of proposed system 14
Fig.9. Binary image 15
Fig.10. Image of digital area 16
Fig.11. Image of the energy meter. 18
Fig.12. Extracted digits of the energy meter reading 18

SICET DEPT.OF EEE PAGE NO:6

 Automatic energy meter reading using image processing

CHAPTER I
INTRODUCTION

In this era of digital world, we have succeeded in automating almost every task, hence reduce the human
effort. Here we would like to introduce Automated Electricity Billing System, used for generating electricity bill
automatically. At present electricity bill is generated manually. An officer from electricity board visit every houses
in a locality and generate the bill by using the meter reading.
This method is not accurate enough. We introduced a new method to generate the bill automatically using
pattern extraction algorithm. Here we use 3 algorithms Canny, Sobel and Prewitt for comparison.
The pattern in the meter is extracted using camera and that image is processed. This image is send to the
Electricity board server using Wi-Fi module. The digit I n the image is used to calculate the bill of the customer.
At present bill generated during every month is not accurate enough. Customers are not able to manage the
usage of electricity. If the unit reading of every month is known to the customer, he can reduce the usage up to a
certain limit. Through this project we can know the current status of the unit and previous 10 months bill.
The existing system is manual. An employee from electricity office visits every house periodically takes the
meter reading and submits this reading at main office. In some houses meter board is inside the house, so meter
readers cannot take reading when no one is in the house. For customers to check the units consumed they have to
look to meter board and it will not be understood by everyone. Calculation may be inappropriate. Some Customers
are not satisfied with existing system. The rest of the paper is organized as follows: Section II contains proposed
system; section III contains module description section IV contains system design and the last sections include
the concluding part of this paper.

SICET DEPT.OF EEE PAGE NO:7

 Automatic energy meter reading using image processing

CHAPTER II
PROPOSED SYSTEM

Proposed system is based on image processing to obtain efficient and accurate reading of the meters. No
human interaction is needed in entire billing process. An app is generated for customers and by log in they can
check current unit consumed and the bill for it. By checking this they can reduce the usage of electricity thereby
reducing the bill generated.

A. Description About the Work

We proposed pattern extraction algorithm for automating the billing system. Pattern extraction algorithm
consist of three main stages of image processing: Preprocessing: It includes Convert RGB to gray scale, Image
binarization, Noise reduction, Crop reading area. The next step is Digit segmentation, then the last step is Digit
recognition, it includes Feature extraction, Template matching. ESP 8266 is the Wi-fi module which we are using.
Wi-fi serial transceiver module. It helps in transmission of the data.ATMEGA328 controller is used. It is 8 bit
processor and is 32 kb.
Using Embedded C language Wi-fi interfacing program is done. At server side calculation of bill is done in
PHP. An app is created for customers. Customer can log in to account using id.

B. Design Description
Image of meter reading is captured from the energy meter. Then we process the obtained reading using 3
pattern extraction algorithm Sobel, Canny and Prewitt. The extracted digit is sent to the electricity board server
using ATMEGA328 micro controller and ESP8266 Wi-fi module

C. Testing and Evaluation

Testing is the process of evaluating a system or its component(s) with the intent to find that whether it satisfies
the specified requirements or not. This activity results in the actual, expected and difference between their results
. In general, given properties are taken under testing phase: Meets the requirements that guided its design and
development, responds correctly to all kinds of inputs, performs its functions within an acceptable time, Is
sufficiently usable.

SICET DEPT.OF EEE PAGE NO:8

 Automatic energy meter reading using image processing

In our project testing is done using sample inputs. The system is already trained with 7 digit segments. But
this is not sufficient for obtaining accurate results. For this we try the system accuracy with the sample inputs. If
the digit is not correct then that particular digit is trained. The error is due to the greater probability while matching
with the data set, hence the wrong character will be taken. This can be reduced by training the data set properly.
This process is repeated till the accuracy improves. By this we can reduce the errors that occur due to segmentation

D. Result and Analysis

We mainly have two sets. Training Set: The System is trained using around 100 samples consisting of digits
from 0 to 9.Training is performed by using OCR Application in MATLAB. Refer fig.1. Each character is trained
with around 4 to 5 samples entered by the user. Further, during testing phase if a digit is not recognized by the
system then it can be trained using OCR application. Testing Set: After the system is trained with the digits, each
digit is tested and if a particular digit is unrecognized then it is added to the training set.

Fig. 1. Training set

SICET DEPT.OF EEE PAGE NO:9

 Automatic energy meter reading using image processing

CHAPTER III

MODULE DESCRIPTION

A. Image Processing Module

We proposed pattern extraction algorithm for automating the billing system. Optical Character Recognition
is used in this project for performing the training of system with digits and for recognizing the digits in the energy
meter. The input entered by user is taken as an image which is given to the OCR for recognizing the digit entered.
In an OCR each character is located and segmented. By preprocessing techniques unique features of character are
extracted to generate training set.

Fig. 2. Output after Image Processing and Digit Recognition

Pattern extraction algorithm consist of three main stages of image processing. This project is done mainly
based on Canny Algorithm. A comparison between the image edge detection methods such as Sobel, Canny and
Prewitt is also been done here by looking the difference in the output we obtain. In our final software we have
options to select the Sobel, Canny and Prewitt methods. We do the project mainly based on Canny algorithm as
it is assumed to be more accurate in edge detection. The common image processing steps can be briefed as follows:
First step is Preprocessing: In the preprocessing step after the image is being captured, it is then re-sized to the
required ratio. We have normally mainly four sub steps in image processing for extracting the digits which include

SICET DEPT.OF EEE PAGE NO:10

 Automatic energy meter reading using image processing

Converting RGB to gray scale, Noise reduction ,Image Binarization and Crop reading area. second step is
Digit segmentation: The proposed idea behind digit segmentation is to scan the cropped numeric area vertically
and horizontally from left to right. This is done by the OCR trainer. The OCR is trained with some inputs of our
previously collected images. Refer fig.2 for output. Third step is Digit recognition: This step have two main
processes which is connected with the digit segmentation process: Refer fig.3. The steps are feature extraction
and template matching.

Fig. 3. Image processing module

B. Arduino Module
After the digit extraction ,the output of the digit recognition program is fed into the Arduino. The process of
sending the extracted digit to Arduino is done by using serial communication. The process takes place only if
there is any value being already recognized by the OCR. Initial settings such as the baud rate which refers to the
speed at which serial communication is done is all been set by default. Our number that is obtained is converted
to string first. It is then sent by opening the port, after sending the data there is a pause of sometime and then the
port is closed. Refer fig.4. Arduino is a micro controller which has Arduino program in it. Here we use Arduino
1.8+.Arduino micro controller is used for sending the extracted digit to the electricity board server

SICET DEPT.OF EEE PAGE NO:11

 Automatic energy meter reading using image processing

Fig.4. Arduino module

C. Website Module
The K.S.E.B office server is now provided with accurate meter reading and it is updated in the database
with the consumer details. The billing is done by converting the power consumption values in Kilo watt hour into
monetary value based on currency exchange rate. Refer fig.5. The billing procedure is followed up by adding tax,
tariff etc. to the amount calculated.
. This allows the consumer to control his electricity consumption. Refer fig.7. The proposed system uses
Android Studio, for implementing the application.

Fig. 5. Website module

SICET DEPT.OF EEE PAGE NO:12

 Automatic energy meter reading using image processing

Fig. 6. Mobile Application module

Fig.7. Output that is viewed by consumer.

SICET DEPT.OF EEE PAGE NO:13

 Automatic energy meter reading using image processing

CHAPTER IV
SYSTEM DESIGN

System design is the process of defining how the information system should be built, ensuring that the
information system is operational and used. It also ensures that the system meets quality standards. At present bill
generated during every month is not accurate enough. Refer fig.8. Customers are not able to manage the usage of
electricity. If the unit reading of every month is known to the customer, he can reduce the usage up to a certain
limit. Through this project we can know the current status of the unit and previous month’s bill.

Fig. 8. System design of proposed system

D. Mobile Application Module

In our proposed system we make use of mobile application for consumers. In the mobile app the consumer
knows the details of the daily consumption of electricity. The consumer can also get the electricity bill and can
pay it online. Refer fig.6. The mobile application also has log in id and password.

SICET DEPT.OF EEE PAGE NO:14

 Automatic energy meter reading using image processing

CHAPTER V
IMAGEPROCESSING
A. Binarization Step
The binary of gray scale images is to use proper threshold to transform the 256 grey scale images into the
binary images, that is, to set the white pixel bigger than the threshold, in the grey scale images, and set black the
pixels smaller than the threshold, to form the binary images with only black and white. Binary images can be
stored in matrix, a, M×N binary image can be expressed by the formula(1), in which i is the line number, j is the
column number, xij is the pixel in line I and column j of I, the binary image , i=1,2,….N.
The threshold is divided into the whole threshold and partial threshold. Here whole threshold is used by
means of iterative. First the 256 gray scale image is counted by means of grey scale histogram, choose the median
of the grey scale image as the beginning threshold V0, and then perform iterative arithmetic as shown in formula
(2) according to h1, the number of pixels with the grey scale value l in I, to obtain the final threshold value V. As
shown in the experiments, when K=0.46, the binary will have the best results. Hence choose K=0.46 to obtain the
binary image (1224 ×1632)

Fig.9. Binary image

B.Cropping Numerical Area

If we count the number of white pixel dots of the binary image following the horizontal line, we will get
the horizontal projection in which the abscissa stands for line number and the ordinate stands for the number of
white pixel dots in response to each of the lines. As the digital area is rectangle, the digits are white, the frame is
black, and the size is fixed, it can be decided that the line number in the digital area is between 400- 600. If the
lines of the image outside the area are cutoff , the horizontal location of the area can be realized.

SICET DEPT.OF EEE PAGE NO:15

 Automatic energy meter reading using image processing

(3) If we count the number of white pixel dots of the binary image following the vertical line, we will get a
vertical projection, in which the abscissa stands for column number and the ordinate stands for number of the
white pixel dots in response to each of the column. As the digital area, with fixed sizes, has the least number of
white pixel dots in the binary image after the horizontal location, it can be decided that the column number in the
digital area is between 250-1380. If the lines of the image outside the area are cut off, the vertical location of the
digital area can be realized and the image of the digital area can be obtained

Fig.10. Image of digital area

C. Character segmentation:
After the digital area has been obtained, it is still necessary to segment the digits in the black frame of the
digital area in order to identify the digits next. In the rectangle black frame of the digital area there are 7 parts
from the left to right i.e. 6 digits and narrow scale division area, with each part in a horizontal line, in response to
the areas a,b,c,d,e,f and g of the vertical projection. As the scale division does not need recognition, it can be cut
off when the digits are segmented. As the characters of the digits have black frame as the background, the
minimum value in a partial area will be obtained in the vertical projection and meet the conditions like the distance
between the characters of the digits. The digital segmentation is performed by means of projection, the background
column between the digits characters , combined with the fixed sizes of the digital characters[6].
Before projection the image has to go through invert dispose. Then the image with reversed color will have
vertical projection so that dots of each column of the black pixels in the digital area will be counted and saved in
array H[j], in which j is the value of columns in the image. For VH=E-D, in which VH is the threshold value, E
is the math expectation and D is the variance in the vertical projection, When H[j]<VH and H[j+1]>0.,then j is
the possible beginning position of the digital characters. When H[j]<VH and H[j-1]>0, then j is the possible ending
position of the digital characters. According to the characteristics of the digital characters the adjacent area after
the beginning position and the adjacent area before the ending position will be checked, adjusted and finally each
of the digits will be segmented according to the adjusted beginning and ending positions.

SICET DEPT.OF EEE PAGE NO:16

 Automatic energy meter reading using image processing

D.Digital Recognition
The energy meter uses 6 digits. Four of the 5 integers usually show whole characters, except that some of
the digital characters ae a little upward while others are a little downward. The last integer and the decimal point
may show the whole characters in the shift of the digits, or show two half characters.
a. Characteristics of the strokes of the digital characters the strokes for the digits 0-9 includes straight
strokes and curve strokes. A straight stroke is a line segment including horizontal, vertical, up-left slanting and
up— right slanting stroke. A curve stroke is a curve segment, including closed curve and non-closed curve. Of
the ten digits, the digit has double closed curves and digits 0,4,6,9 have single closed curves. The digits 1,2,3,5,7
have non closed curves. If the closed curve of the single closed curve digit is at the upper part, it is 9. If it is at
lower part it is 6 or 4; or if it is in the middle it is 0. 4 and 6 can be discriminated by means of vertical stroke. The
digit 1 has only a vertical stroke and the digits 2,3 and 7 have horizontal stroke. The horizontal stroke of 2 is at
the bottom. 7 has up-left slanting strike besides a horizontal stroke. 5 has both horizontal and vertical strokes.
b. Digital Recognition Algorithm This algorithm involves three steps,
Step 1: Determine datum line of the digits. For this make a horizontal projection of the digit and count the
dots of the black pixels in each line. If there is a blank area at both ends of the projection, the datum line is fixed
on the center of the bigger blank area in response to the whole character. If the blank area is only in the middle of
the projection, the datum line is fixed on the center line of the blank area in response to the half character.
Step 2: Search for the closed curve area of the digital characters. For this choose a white pixel dot as A and
search for an area connected to A and mark all the pixel dots in the connected area as background1. If all the white
pixel dots in the image have been marked as background 1, this shows that that character is a non-closed curve
character and the search can be stopped. If some white pixel dots in the image have not been marked as background
1, then choose one of the dots as B and continue the search for the area connected to B and mark all white pixel
dots in the connected area as background 2. If all the white pixel dots of the image have been marked as
background 1 and background 2, this shows that the character is a single closed curve digit. Otherwise , it is a
double closed curve digit.
Step 3: According to the datum line, the digits are divided into 4×2=8 smaller areas. According to search
result of the closed curve area, combined with the characteristics of the related digital strokes , search in each
smaller area and finally determine the digital character.

SICET DEPT.OF EEE PAGE NO:17

 Automatic energy meter reading using image processing

CHAPTER VI
RESULTS
With the help of above image processing technology digits are extracted from the digital are of the image
taken by camera and transferred to the server PC.

Fig.11. Image of the energy meter.

Fig.12. Extracted digits of the energy meter reading

SICET DEPT.OF EEE PAGE NO:18

 Automatic energy meter reading using image processing

CHAPTER VII
CONCLUSION
This paper presented automatic electricity meter reading system based on image processing. Comparison
between various edge detection methods were also presented here. The proposed system shows that rather than
manual procedure of taking electricity meter reading, we can implement a process in which the meter reading is
captured with the help of a image capturing device.

By applying the Edge Detection technology on captured image, we can extract the reading from the image
and it is sent to the electricity board with the help of a network connection. The user can also track their daily
electricity usage with help of an application that can be installed and used. This idea reduces the difficulty of the
employee and increases the efficiency. Thus while implementing it would be at low cost with the desired results
as the consumer wishes.

In future we can implement this project without using camera. So we can reduce the implementation cost.
The digit in the meter can be send to the server directly without fixing a camera in front of the energy meter.

SICET DEPT.OF EEE PAGE NO:19

 Automatic energy meter reading using image processing

CHAPTER VIII
REFERENCES

[1] Md. Masudur Rahman; Noor-E-Jannat; Mohd. Ohidul Islam Md. S. Salakin“Arduino and GSM based
smartenergy meter for advanced metering and billing system” 2015 International Conference on Electrical
Engineering and Information Communication Technology(ICEEICT)
[2] Chinu; Amit Chhabra ;“A hybrid approach for color based image edge detection” 2014 International
Conference on Advances in Computing, Communications and Informatics (ICACC
[3] Kanchan Mahajan; Deepali Jaybhave; Namrata Nagpure; Bhagyshri Shirsat;”A novel method for automatic
electricity billing system us ing Ad-Hocnetworks “
2016 International Conference on Global Trends in Signal Processing, Information Computing and
Communication (ICGTSPICC
[4] C. Edward, Support Vector Machine Based Automatic Electric Meter Reading System, IEEE International
Conference on Computational Intelligence and Computing Research, 2013, pp. 1-
[5] K. Parthiban and A. M. Palanisamy, Reading Values in Electrical Meter Using Image Processing
Techniques, IEEE International Conference on Intelligent Interactive Systems and Assistive Technologies,
Coimbatore, INDIA, 2013, pp. 1-6
[6] S. Zhao, B. Li, J. Yuan, G. Cui, Research on Remote Meter Automatic Reading Based on Computer Vision,
IEEE/PES Transmission and Distribution Conference, China Dec, 2005, pp. 1–4
[7] M. Shinde and P. Kulkarni, Reading of Energy Meter based on Image Processing Technology, International
Journal of Electronics Communication and Computer Engineering, Volume 5, Issue (4) July, Technovision-
2014, pp. 1-5.
[8] R. F. Mansour, A Robust Method for Arabic Car Plates Recognition and Matching Using Chain Code,
American Journal of Computational and Applied Mathematics 2012, 2(3): 105-111
[9] Granger, H., Chanchad, R., & Razak, S., License Plate Number Detection Algorithm for Qatari License
Plates, 2012, pp.1-6.

SICET DEPT.OF EEE PAGE NO:20