ISSN 2347 - 3983

Anusha K et al., International Journal of Emerging Trends in Engineering Research, 9(8), August 2021, 1107 – 1110
Volume 9. No. 8, August 2021
International Journal of Emerging Trends in Engineering Research
Available Online at http://www.warse.org/IJETER/static/pdf/file/ijeter13982021.pdf
https://doi.org/10.30534/ijeter/2021/13982021

Indian Currency Detection Using KNN Classifier

Anusha K1, Dr. Sreedharamurthy S K2
1
PG Student, Department of Electronics and Communication, UBDTCE,
Davangere,Karnataka,India,[email protected]
2
Professor and Head of Department, Department of Electronics and Communication, UBDTCE,
Davangere,Karnataka,India,[email protected]


ABSTRACT
2.2 IMAGE PREPROCESSING
Fake Currency Detection is the biggest problem faced by
many countries including India. The advancement of colour Image Acquisition: the images are obtained through webcam
printing technology has increased the rate of fake currency or digital camera, they were further processed using
note printing and duplicating the notes on a very large scale. MATLAB to extract the information required.
So, it has become a necessity to develop a tool that detects the
fake currency note in a less time and in a more efficient Image pre-processing : Image pre-processing is to suppress
manner using Image processing. undesired distortions or enhance some image features that are
important for further processing or analysis.
Key words : Log Gabor, Contrast enhancement, KNN
Classifier, Image acquisition, Image Processing. Image Adjusting : During image adjusting, after getting the
image from a web camera, the size of the image should be
1. INTRODUCTION reduced as per requirement and it is done with the help of
image interpolation.
In recent years a lot of fake currency note is being printed
which have caused great loss and damage towards society. Image Smoothening: During Image Smoothening, it is
With the production and circulation of fake notes, it becomes necessary to remove the noise that may affect segmentation
difficult for common people to differentiate whether the and pattern matching. Mask values can be used to determine
currency is real or fake as they differentiate on the basis of the degree of smoothing.
physical appearance that may fail many times. In this paper,
we propose a system for Verification of currency note using RGB to Gray scale conversion: The image acquired is in
image processing techniques and KNN Classifier. The RGB colour. This image is heavy and has more noise. By
proposed system has got advantages like simplicity and high converting it into gray scale, reduces the size of the image and
also the intensity information which is easy to process instead
performance speed. The result will predict whether the
of processing three components R (Red), G (Green),
currency note is fake or real. The system has more efficient
B(Blue).Gray scale Images Representation is the way of
performance of and achieves good accuracy rate. [2][4]
representing an image only by using different shades of gray
levels. Gray level image requires less information to be
2. IMAGE PROCESSING WITH MATLAB provided for each pixel.
Until now, many techniques have been proposed to identify Edge detection: It is a set of mathematical methods which
the currency note. If a note is dirty, its color characteristics are aim at identifying points in a digital image at which the image
changed widely. So it is important that how we can extract the brightness changes sharply or, more formally, has
features of the image of the currency note and apply proper discontinuities. The points at which image brightness changes
algorithm to improve accuracy to recognize the note. sharply are typically organized into a set of curved line
Here, the image is captured by a simple digital camera such segments termed edges. It is a fundamental tool in image
that all the features are highlighted. Image is then stored for processing, machine vision and computer vision, particularly
further processing. in the areas of feature detection and feature extraction.

2.1 IMAGE CAPTURE: Web Camera is used to capture the

image which is an input image given to Matlab for processing.
A webcam captures and it sends the data to the computer.

1107
Anusha K et al., International Journal of Emerging Trends in Engineering Research, 9(8), August 2021, 1107 – 1110

3. IMPLEMENTATION OF THE SYSTEM Figure2 shows the classification of two groups using KNN
classifier. K is a constant factor which is the value obtained by
taking square root of number of samples taken for training.
KNN Algorithm uses feature similarity to predict the values
of new data points which further means the new data point
will be assigned a value based on how closely it matches the
points in the training set. Working of the algorithm is
explained by following steps:
Step1- for implementation of algorithm, load the training as
well as data set.
Step 2-Choose the K value, which is any integer value.
Step 3-For each point in the test data do the following:
3.1- calculate distance between test data and each row of
training data using Euclidean distance method.
3.2- Now, based on the distance value, sort them in
ascending order.
3.3- Next, it will choose the top K rows from the sorted
array.
3.4- now, it will assign a class to the test point based on
most frequent class of these rows.
Step 4- End.

B. Gamma Correction: Gamma correction controls the

overall brightness of an image. Each pixel in an image has
brightness level called luminance having the values between 0
to 1, where 0 means complete darkness (black) and 1 means
brightest (white). Gamma correction is a technique to
compensate for non-linear display characteristics of a device
by mapping input values through a correction function
corresponds to the characteristics of display device. It is done
by sending signal through inverse function .
Figure 1: Implementation of the system.
C. Contrast Enhancement: Contrast is defined as the
difference in intensity between two objects in an image.
3.1 TERMINOLOGIES USED IN THE PROPOSED Histogram represents the number of pixels for each intensity
value considered. Histogram equalization is one of the
SYSTEM contrast enhancement technique widely used in image
A. K-Nearest Neighbor (KNN) Classifier: processing because of its high efficiency and ease of use. This
modifies dynamic range and contrast of an image so that the
resulting image histogram has the desired shape.

D. Log-Gabor Filter: Log Gabor is a linear filter used for

texture analysis. It is a logarithmic function, by varying the
values of Gabor coefficients – x from 0 to 255 in the
exponential value we need to create a matrix having the same
order as that of input image matrix. The obtained matrix is
known as Gabor filter which is to be convolved with the gray
scale matrix to get the Gabor features of the input image. The
obtained Gabor feature values are normalized in the range of
0-1. If the values are 0, it represent blue and if it is nearer to
zero it represent green, if the values are nearer to 1 it represent
yellow and if the value is 1 then it represent red colour as seen
and numbered as 1d and 2d in figure 3and 4respectively.

Figure 2: KNN-Classifier for two testing groups.

1108
Anusha K et al., International Journal of Emerging Trends in Engineering Research, 9(8), August 2021, 1107 – 1110

4. RESULTS AND DISCUSSION

120 images of fake and real Indian currency images of

denomination Rs.10 and rs.20 are used for processing. After
preprocessing, features coefficients are obtained by applying
Log-gabor filter. These features are in turn given to KNN
classifier for recognition. Satisfactory recognition rate is
obtained after implementation. Results are given as below.
Figure3 shows the outcomes of testing Real Rs.10 Note and
Figure4 shows the outcomes of testing Fake Rs.10 Note.

Figure 4: Outcomes of testing Rs.10 Fake note.

In this system of recognition of Indian currency notes, we

have used 120 Images of real and fake Currency notes of
denomination of Rs.10 and Rs.20 for Classification. Out of
which 60 images (Rs.10 & Rs.20) are used for training the
system and 60 images (Rs.10 & Rs.20) are used for testing the
system. For this process we have used KNN classifier. The
obtained results are tabulated below.
Figure 3: Outcomes of testing Rs.10 Real note.

1109
Anusha K et al., International Journal of Emerging Trends in Engineering Research, 9(8), August 2021, 1107 – 1110

5. Vivek Sharan, Amandeep Kaur, Detection of

Counterfeit Indian Currency Note Using Image
Processin, (IJEAT) ISSN: 2249 – 8958, Volume-9,
Issue-1, October 2019.
6. M.Deborah and Soniya Prathap Detection of Fake
currency using Image Processing. IJISET-International
Journal of Innovative Science, Engineering &
Technology, Vol. 1, Issue 10, 2014.
7. Image processing toolbox help, MATLAB® [Online].
Available: http://www.mathworks.com/
8. Digital Image Processing Using MATLAB – By Rafel C.
Gonzalez & Richard E. Woods (Addison Wesley)
9. D.Alekhya , G.Devi Surya Prabha , G.Venkata Durga
Rao, Fake Currency Detection Using Image
Processing and Other Standard Methods,
International Journal of Research in Computer and
Communication Technology, Vol 3, Issue 1, January-
2014.
Table 1: Percentage of recognition rate of the system.
10. Prof. Renuka Nagpure, Shreya Shetty, Trupti Ghotkar,
Chirayu Yadav, Suraj Kanojiya, Currency Recognition
and Fake Note Detection, International Journal of
Innovative Research in Computer and Communication
5. CONCLUSION Engineering (IJIRCCE), March 2016.
Indian currency notes of denomination Rs.10 and Rs.20 are
used in our system for classification. After preprocessing
these images, features are extracted by applying the
Log-gabor filter. Filter coefficients are then given to KNN
classifier for recognition purpose.The features are compared
with trained dataset features Using KNN Classifier. We have
used 120 Images of real and fake Indian Currency notes of
denomination of Rs.10 and Rs.20 for Classification. For
training the system we have used 60 images, that includes 30
images of denomination Rs.10 and 30 images denomination
Rs.20. For testing the classifier we have used 60 images that
also includes 30 images of denomination Rs.10 and 30 images
denomination Rs.20. The percentage of recognition rate is
calculated based on the rate of correct classification of test
image as real and Fake Notes. The Table1 represents the
percentage of recognition rate of the system . The average
recognition rate of the system is 92%. Hence the proposed
system is having more efficient performance and achieves
good recognition rate.

6. REFERENCES
1. Y. Neeraja, B. Divija and M. Nithish Kumar Fake
Currency Detection Using K-nn Technique,
IJREITSS, Vol. 09, No. 1, pp. 201-205, 2019.
2. P. D. Deshpande and A. Shrivastava, Indian Currency
Recognition and Authentication using Image
Processing , IJARSE, Vol. 07, No. 7, pp. 1107-1119,
2018.
3. Saiyed Mohammed Arshad, Devdatt Sawant Sudagar &
Nausheeda B S3, Fake Indian Currency Detection
Using Image Processing, IJLTET, Special Issue
SACAIM 2017, pp. 598-602 e-ISSN:2278-621X.
4. Devid Kumar, Surendra Chauhan, Indian Fake
Currency Detection Using Computer Vision,
IRJET,Volume: 07 Issue: 05 | May 2020.

1110