B.L.D.E. A’S V.P. Dr. P.G.
HALAKATTI COLLEGE OF ENGINEERING AND
TECHNOLOGY BIJAPUR – 586 103
DEPARTMENT OF
COMPUTER SCIENCE AND ENGINNERING
PROJECT REPORT ON
“CURRENCY RECOGNITION SYSTEM USING IMAGE PROCESSING”
Submitted in partial fulfillment for the award of
BACHELOR OF ENGINEERING
IN
COMPUTER SCIENCE AND ENGINNERING
UNDER THE GUIDANCE OF
Prof. KIRAN B PATIL
SUBMITTED BY
NAME USN
1.VIRESH MATH 2BL16CS105
2. ABHILASH BALLARI 2BL17CS400
3. SUMEET POL 2BL17CS412
B.L.D.E. A’S V.P. Dr. P.G. HALAKATTI COLLEGE OF ENGINEERING AND
TECHNOLOGY BIJAPUR – 586 103
DEPARTMENT OF COMPUTER SCIENCE AND
ENGINEERING
CERTIFICATE
This is to certify that the project work Phase I entitled “CURRENCY RECOGNITION SYSTEM
USING IMAGE PROCESSING” is a bonafide work carried out by VIRESH MATH, ABHILASH
BALLARI,AND SUMEET POL (2BL16CS105,2BL17CS400,2BL17CS412)
and Submitted in partial fulfillment for the award of Degree of Bachelor of Engineering
th
in VII Semester of Visvesvaraya Technological University, Belgaum during the year
2019-2020.
GUIDE HOD PRINCIPAL
Prof. Kiran B Patil Dr. Pushpa Patil Dr. V.P. Huggi
ACKNOWLEDGEMENT
We would like to express deep sense of gratitude to our Principal Dr. V. P. HUGGI for
providing all the facilities in the college.
We would like to thank our Head of the Department Dr. PUSHPA PATIL for providing all the
facilities and providing a good academic environment in the department.
We feel deeply indebted to our esteemed guide Prof. KIRAN PATIL for the help, right from the
conception and visualization to the very presentation of the project. He has been our guiding
light throughout.
We would like to take this opportunity to thank all the faculty members and supporting staff for
helping to do this project.
Last but not the least we would like to thank Almighty and our parents, friends and well-wishers
who have helped us directly or indirectly.
VIRESH MATH
ABHILASH BALLARI
SUMEET POL
CONTENTS:
CHAPTER 1: ABSTRACT ---------------------------------------------------------------- 01
CHAPTER 2: INTRODUCTION ---------------------------------------------------------02
CHAPTER 3: LITERATURE SURVEY--------------------------------------------------04
CHAPTER 4: MOTIVATION--------------------------------------------------------------07
CHAPTER 5: OBJECTIVES AND PROBLEM DEFINITION------------------------08
CHAPTER 6: ARCHITECTURE OF PROPOSED SYSTEM--------------------------09
CHAPTER 7: TOOLS AND TECHNOLOGY--------------------------------------------12
CHAPTER 8: APPLICATION -------------------------------------------------------------13
CHAPTER 9: IMPLEMENTATION------------------------------------------------------14
CHAPTER 10: CONCLUSION ------------------------------------------------------------15
1.ABSTRACT
we propose a system for automated currency recognition using image processing techniques. The
proposed method can be used for recognizing both the country or origin as well as the
denomination or value of a given banknote .
Only paper currencies have been considered. This method works by first identifying the country
of origin using certain predefined areas of interest, and then extracting the denomination value
using characteristics such as size, color, or text on the note, depending on how much the notes
within the same country differ.
We have considered 20 of the most traded currencies, as well as their denominations. Our system
is able to accurately and quickly identify test notes.
1
2.INTRODUCTION
According to the survey conducted by the CIA , there are around 180+ currencies presently
circulating in the world. Each of these currencies differs greatly in features such as size, color
and texture. Unlike the olden times, the trade and commerce between countries has increased
in all sorts of levels.
The need for acquiring knowledge about all the currencies by the banks has been extremely
important. However for any human teller to recognize each note correctly is something that is
not feasible. Thus the need for an efficient automated system that helps in recognizing notes is
pivotal for the future.
All currencies around the world look totally different from each other. For instance the size of
the paper is different, the same as the color and pattern. The staffs who work at places like
money exchange offices have to distinguish between different types of currencies and that is
not an easy job. They have to remember the symbol of each currency. This may result into
wrong recognition, so they need an efficient and foolproof system to aid in their work.
The aim of our system is to help people who need to recognize different currencies, and work
with convenience and efficiency. With development of modern banking services, automatic
methods for paper currency recognition become important in many applications such as
vending machines.
It is very difficult to count different denomination notes in a bunch. This project proposes an
image processing technique for paper currency recognition and conversion. The extracted
region of interest (ROI) can be used with Pattern Recognition and Neural Networks matching
technique.
This project is basically an idea to design system which is used for currency recognition . Each
country has its own different currency so it is very complicated task for people to recognize the
currency.
In manual currency recognition system , there are many problems. We will be developing this
system to overcome those problems which have been faced.
2
List of chosen countries:
SR.No Country
1 Australian Dollars
2 Canadian Dollars
3 Chinese Renminbi
4 American Dollars
5 Euro
6 Hong Kong Dolla
7 Indian Rupee
8 Indonesian Rupiah
9 Kuwaiti Dinar
10 Mexican Peso
11 New Zealand Dollar
12 Philippine Peso
13 Japanese Yen
14 Russian Rubles
15 Saudi Riyal
16 Singapore Dollar
17 UAE Dirhams
18 Swiss Franc
19 Norwegian Kroner
3
3. LITERATURE SURVEY
1. J. Guo et al. in 2010 proposed an improved LBP algorithm, called block-LBP algorithm, which
is proposed for characteristic extraction. The proposed method has advantages of simplicity
and high speed. The experimental results show that this improved method has a high
recognition rate, as well as robustness for noise and illumination change. There have been no
researchers who applied LBP feature to automatic recognition of paper currency before them.
The experimental results have shown effectiveness of the proposed method.
2. B. Vanajakshi, D. Gowthami and N. Mounika in 2017 proposed algorithm uses the
primary color and a part of currency in hsv components by fixing the saturation and value
threshold levels for recognition. In future the on basis of image acquisition, multiple parameters
including correlation matrix, edge detection operators, color check etc were also considered.
3. Devashree R. Patwardhan, Swarali Namjoshi, Vrinda Valunj, Pratibha Shikhare, Prof. Anjali
Shejul in 2017 proposed algorithm that solved the major issues related to currency
recognition. One of the initial feature of this system is obtaining the image and it basically
focuses on an image that can be obtained by using number of different equipment’s, such as
cameras or Scanner. One approach is basically based on conversion of RGB value into HSV value
and they also maintain the distance of each image by calculating Euclidean distance formula
and then compare the distance with test image which give accurate result. Also provide the
user with detail information of currency in PDF format and especially for the blind people we
provide the facility of voice generator.
4. Prof. Sagar S.Rajebhosale, Devang S.Gujarathi, Sushil V.Nikam, Prathmesh P.Gogte,
Nilesh M.Bahiram in 2017 proposed system is based on image processing and makes the
process automatic and robust. Shape information are used in there algorithm. Original Note
Detection Systems are present in banks but are very costly. they are developing an image
processing algorithm which will extract the currency features and compare it with features of
original note image. This system is cheaper and can provide accuracy on the basics of visual
contents of note. So, as an output, people will get information provided the note image is
original or duplicate. It is accurate and highly-efficient. But for most staffs,they have to keep a
lot of different features and anti-fakes label for different commonly-used currency in their
mind.
5. Rubeena Mirza, Vinti Nanda in 2012 proposed , Design and Implementation of Indian Paper
Currency Authentication System Based on Feature Extraction by Edge Based Segmentation
Using Sobel Operator. Three characteristics of Indian paper currency is selected for counterfeit
detection included identification mark, security thread and watermark. The characteristics
4
extraction is performed on the image of the currency and it is compared with the characteristics
of the genuine currency. The sobel operator with gradient magnitude is used for characteristic
extraction. Paper currency recognition with good accuracy and high processing speed has great
importance for banking system. The proposed method has advantages of simplicity and high
speed. The experimental results show that this approach is effective and efficient and clearly
meet the system requirements.
6. B. V. Chetan et al. in 2012 proposed the side of the precious paper currency recognition
method. It is a two-stage: 1. Matching database notes Matching using correlating the edges of
input. The results showed that the method of identification of paper currencies, "Gabor
Muweijeh", an accuracy of 65%. the method of subtraction image gave a resolution of 51.52%,
and the method based on the Local Binary Pattern (LBP) gave accuracy of 52.5%. While the
suggested technique gave very high rate exceed to ninety nineand half percentage accuracy for
a particular set of currency .
7. A. Rajai et al. in 2012 introduced a method to extract the texture profiles of the currency
memo pictures. Using Discrete-Wavelet-Transform (DWT) with a group of statistical measures
extracted from the approximate.
8. A. B. Sargano et al. in 2014 proposed smart systems to recognize the Pakistani-paper
currency. After finding the features, they suggest to use multi-layers of forward feed
Propagation Neural Network (PNN) for classification. The technique was simple and consumed
relatively less time making them suitable for real-time application. The results indicated that
the system had the ability to recognize 99% efficient of captured images .
9. K. Vora et al. in 2015 proposed an algorithm based on the method of extracting the
frequency band feature and discussed the currency detector. 2Dimensional "Moji" Discrete
Wavelet Transform (2D-DWT) and a group of statistical moments used. Extracted features could
be used to identify, classify and retrieve bank notes. The result of the classification would
facilitate the recognition of the counterfeit currency based on the serial number extraction
mainly through the implementation of OCR .
10. S. Sahu and et al. in 2016 reported an image processing technique used to recognize the
paper currency of different countries. A booming approach to determine the paper currency
depending on the pre-processing, extracting feature and classification of those currency images
was discovered. The step for the currency identification system of different processing steps
for analyzing the definition of paper currency were pre-treatment, morphological filtration,
analysis, induction analysis, fragmentation and extraction of properties.
5
11. Singh et al. in 2011 proposed the heuristic analysis of Indian currency notes and the digits
of serial number of Indian currency notes to recognize them. To identify a character of currency
image the features of that image should be extracted. It is very important to extract features
from different notes. To extract correct features of character heuristic analysis are done before
extracting features in currency recognition.
12. Anggarjuna Puncak Pujiputra, Hendra Kusuma, Tri Arief Sardjono in 2018 proposed an
ultraviolet (UV) Rupiah paper currency image recognition by implementing Gabor wavelet
feature extraction. The UV image is used to distinguish between a genuine and a fake paper
image currency, since under UV light a different visual in specific areas of the real banknote will
glow and show hidden patterns. To have a high accuracy as well as efficiency, we use 3 scales
and 8 orientations Gabor bank and subspace-LDA classifier in recognition process. The
proposed Gabor method has advantages of easiness and high accuracy. The experimental
results demonstrate that this method is quite reasonable in terms of preciseness, with 98.5%
overall average recognition rate are obtained for the data of 160 UV Rupiah paper currency
images.
6
4.MOTIVATION
Around 180+ currencies are available around the world and the need for an automated system
related to currencies has been increasing exponentially recently. The need for developing
systems that process notes without human intervention for various different uses has been
pivotal for the development of systems that help in detecting and recognizing currency notes.
However the varying features in each notes and the security aspects involved in different
currencies make this task extremely difficult. Detection of fake currencies which is spreaded in
Indian market also our main goal is to use image processing technique.The proposed web portal
will help common people for currency recognition anywhere anytime. Automatic method for
detection of fake currency note is very important in every country.
7
5. PROBLEM DEFINATION AND OBJECTIVE
5.1 Problem Definition
This project is basically an idea to design system which is used for currency recognition .Each
country has its own different currency so it is very complicated task for people to recognize the
currency. In manual currency recognition system , there are many problems. We will be
developing this system to overcome those problems which have been faced.
5.2 Objectives
1. The main objective is to identify the country of origin and denomination (value) of the note.
2. Various notes of different value (denomination) from the same country generally can be
differentiated based on certain features such as size ratio, color, or in the worst case, a text
extraction to obtain the value from the note itself.
3. The need for an efficient automated system that helps in recognizing notes which is pivotal
for the future.
4. In order to differentiate the currency based on two different parameters (country of origin,
and denomination), we segment the problem into two steps:
1.First, identify the country of origin .
2. Identify the denomination (value) of the note.
The reason we chose this approach is based on the observation that various notes of different
value (denomination) from the same country generally can be differentiated based on certain
features such as size ratio, color, or in the worst case, a text extraction to obtain the value from
the note itself.
8
6. ARCHITECTURE OF PROPOSED SYSTEM
In order to differentiate the currency based on two different parameters (country of origin, and
denomination), we segment the problem into two steps:
1.First, identify the country of origin .
2. Identify the denomination (value) of the note.
The reason we chose this approach is based on the observation that various notes of different
value (denomination) from the same country generally can be differentiated based on certain
features such as size ratio, color, or in the worst case, a text extraction to obtain the value from
the note itself.
The full method has been described briefly below:
A. Pre-processing
The image of the banknote must first be pre-processed to remove any extraneous noise. This
is done by applying a simple de-noising filter. The image is then converted to a binary image
using adaptive thresholding. This allows us to identify the empty regions of the banknote as
those with all black pixels. Note that these empty regions correspond to regions that are free of
any foreground objects in the actual banknote. Although there may be some background
patterns, these will be removed by de-noising followed by converting to a binary image. The
image is also resized to allow for easier computations.
B. Identify the country of origin
1) Identifying empty regions:
Once the pre-processing steps have been done, we can identify which regions of the note are
relatively empty (black pixels in the binary image). This is done based on certain predefined
areas. All the currencies are clustered into groups based on which regions of the note are
relatively empty. We have chosen to divide them into 3 groups – left side empty, right side
empty, and center empty, although if the number of currencies were larger, we could possibly
use a larger set of groups (top empty, bottom empty, etc.).
9
Grouping is done by finding out the ratio of black to white pixels for the required region, and
then classifying the note based on this ratio. The values chosen to classify the notes have been
found experimentally. Note that some notes have no significant empty space, and therefore
don’t fall into any of the groups. These notes are classified into another group.
2) Using template matching to identify country of origin:
Once the banknote has been segregated into one of the predefined groups, we can check the
image against templates for each of the countries within that group. Note that this is requires
less comparisons than checking the image against all templates of every country in the system,
and is the reason we have chosen to segregate the countries into such groups. The templates
are chosen such that they are small (thus requiring less computation) but still uniquely identify
the country of origin. Thus, the templates are chosen to be uniform symbols such as the
country’s seal, name of the country itself in stylized font, etc. We can also template match the
given banknote in only the region that we know a certain symbol will be, if the location is
uniform across all denominations. For example, the maple leaf symbol on the Canadian Dollar
can be used as a template, as its location is uniform across all Canadian notes (top left corner).
Thus, we can template match for the Canadian maple leaf in only this section of the note. This
reduces the amount of computation.
C. Identify the denomination
Once the country has been correctly identified in the previous step, we can try to identify the
value of the note. There are three different approaches used for this purpose:
• Size ratio
• Color
• Text extraction
The methods are listed in order of increasing computation time. Some countries have
banknotes that can be easily differentiated by the size. If the country of origin has been
identified as one of these, then we can simply compare the size of the given banknote with the
known sizes of all the denominations. In some countries, the banknotes are of too similar size
to differentiate them based on this feature alone. In that case, we can try using difference in
color between the different denominations. For this, the k-means clustering algorithm is run on
the given banknote to extract the dominant color. This is done on the image in LAB color space.
10
LAB color space consists of 3 channels: the L channel corresponds to lightness value, and the a*
and b* channels represent the color value. LAB color space provides a more accurate
perception of color. Once the value of the dominant color is obtained, we compare it with the
known color values of all the denominations of that country. The denomination that has the
least square distance of the a* and b* channels is the one with the least color difference, and is
selected as the actual denomination of the note. If both the above methods fail, then we must
extract the value of the denomination from the note using text extraction. As this is the most
computationally intensive method, we only apply this is case of countries such as USA, etc.
where both the color and size does not vary too much between the denominations. If both the
other methods do not work, this method is sure to work, as all banknotes have the value of the
denomination written on the note in at least one place. Thus, any of these three methods works
for every note in the set of currencies taken.
Figure 6.1: Block diagram of proposed system
11
7. TOOLS AND TECHNOLOGY
MATLAB:
When it comes to image processing what can be more flexible than MATLAB. It stands for
“Matrix Laboratory”, a fourth-generation high-level programming language developed by
MathWorks (U.S). It was created in the year 1984 with an objective to provide interactive
environment for computation, visualisation and programming. It is written in C, C++ and Java.
MATLAB is used in every facet of computational mathematics. Mathematical calculations where
it is most commonly used include- matrix and array manipulations, linear algebra, algebraic
equations, statistics, calculus, integration, transformation, etc. It has wide range of applications
including signal processing, image and video processing, control systems, computer vision, AI,
etc. MATLAB is the most popular software used in the field of Digital Image Processing.
However, it is not open source, a user has to pay for licensed MATLAB interpreter.
CURRENCY IMAGES:
currency image taken from mobile.
SOFTWARE AND HARDWARE REQUIRED:
MATLAB 9.7 application
4GB RAM
20GB internal space
12
8. IMPLEMENTATION
To identify the country of origin, we have first segmented the note into a group based on presence
or absence of empty regions in certain pre-defined areas. Then, we classify the note by doing
template matching with templates that characterize the banknotes from each of the countries in the
group. For demonstration purposes, the Canadian 20 dollar bill is chosen.
Once the image has undergone preprocessing, to identify the country of origin, we template
match it against the templates of all the countries within its group. For Canadian notes, we have
chosen to use the maple leaf that is present in the top left corner as a template, as it is present in
all the Canadian notes, and is a simple enough image.
Once the country has been determined, the denomination should be identified. This can be based
on color, size, or text extraction. The color of the note can be used to differentiate notes such as
Canadian Dollar, Mexican Pesos, etc., as they vary in color significantly between different
denominations.
Size can be a differentiating factor when the different denominations are of significantly different
size. This can be used for Swiss Franc, etc. The different denominations of the Swiss Franc and
their relative sizes are for comparison.
We used Tesseract to train the system with the necessary information. The US 2 dollar note is
taken as example. We resize it, and then de-noise it so that background noise is removed. This is
essential to extract clean text. The image is then cropped to the area which holds the text. The text
reads the denomination of the note in words.
We apply adaptive threshold, followed by bit-inverse to the cropped image. The adaptive
threshold converts the image to only Black/White form. The bit-inverse interchanges the black
and white bits, so that we get black text against white background.
It consequently detects the characters and gives the desired output. The average time for
determining the value of the various banknotes has been tabulated in the Table II. Overall, we
have found that our system is able to accurately recognize most of the countries and
denominations correctly (93.3% accuracy, where accuracy is defined as the number of notes
correctly identified divided by the total number of notes tested). This is a much better result than
that of the crude algorithm, which fails to recognize more than half of the given test images
accurately. The crude algorithm that we used did a brute force comparison of the pixel by pixel
mean square distance for each image.
13
9. APPLICATIONS
1. This system is used to identify the currency note in businesses, banks, malls, railways,
organizations, etc. but it is mainly recognized using a hardware device. Common man also
doesn’t always have the ability to use it as a hardware. Therefore, there is a need to
computerize the human effort to recognize the currency.
2. System compare images of currency note to the stored images of original currency note
images.
3. To provide Cheaper and Accurate system to the user which can easily accessible and gives
accurate recognition of currency notes.
4. To develop user friendly web application of currency recognition system.
5. To make available to common people quickly and easily so they can utilize anywhere and
at any time.
6. Despite the quickly expanding utilization of master cards and other electronic types of
payment,money is still broadly utilized for ordinary.
14
10.CONCLUSION
In conclusion, we will be designing a system that accurately identifies both the country of origin
and the denomination of a given banknote. Our system currently supports twenty of the most
common currencies, but can easily be extended to more countries based on the method we have
previously described. When compared with the crude algorithm of pixel by pixel comparison, our
algorithm is considerably more accurate, and takes less time. We have thus learned that our
proposed algorithm is able to identify currency and denomination in an average of 5.3 seconds,
which is a considerable improvement over the crude algorithm. However, our proposed system only
considers a limited number of currencies. There are 180+ currencies that can be included in the
system, and we have chosen to only do for 20 of the most common ones. Also, the system should
be effective in identifying notes that are mutilated. Our system is not effective under this
consideration.
15
REFERENCES
[1] Debnath, K.K., S.U. Ahmed and Md. Shahjahan, 2010. A paper currency recognition system
using negatively correlated neural network ensemble. J. Multimedia, 5(6): 560-567.
[2] Qing, B. and J. Xun, 2010. Currency recognition modeling research based on BP neural network
improved by gene algorithm. Proceeding of the 2nd International Conference on Computer Modeling
and Simulation, pp: 246-250.
[3] Jahangir, N. and A. Raja, 2007. Bangladeshi banknote recognition by neural network with axis
symmetrical masks. Proceeding of the 10th International Conference on Computer and Information
Technology, pp: 105.
[4] Guo, J., Y. Zhao and A. Cai, 2010. A reliable method for paper currency recognition based on
LBP. Proceeding of the 2nd IEEE International Conference on Network Infrastructure and Digital
Content, pp: 359-363.
[6] Singh, P., G. Krishan and S. Kotwal, 2011. Image processing based heuristic analysis for
enhanced currency recognition. Int. J. Adv. Technol., 2(1): 82-89.
[7] Hanish Agarwal and Padam Kumar, ”Indian currency note Denomination recognition using color
images”, International Journal of Computer Engineering and Communication Technology, ISSN
2278-5140, Vol 1, Issue 1.
[8] Dipti Pawade, Pranchal Chaudhari and Harshada Sonkamble, “Comparitive Study of Different
Paper currency and coin currency Recognition Method”, International Journal of Computer
Application, ISSN 0975-8887, Vol 66, No 23, Mar 2013.
[9] Faiz M. Hasanuzzaman, Xiaodong Yang, and YingLi Tian, “Robust and Effective Component-
based Banknote Recognition by SURF Features”.
[10] Rumi Ghosh and Rakesh Khare,” An Elegant Neural Network based draw near for currency
Recognition”, Journal of Environmental Science, Computer Science and Engineering and
Technology, JECET; June – August-2013; Vol.2.No.3, 876-882.
[11] Pragati D Pawar and Shrikant B. Kale,” Recognition of Indian Currency Note Based on HSV
Parameters”, ISSN 2319-7064, Vol 3, Issue 6, Jun 2014.
[12] John R. Smith and Shih-Fu Chang, “Tools and Techniques for Color Image Retrieval”, IS&T/SPIE
Proceedings Vol. 2670.
16
