DETECTION OF INDIAN COUNTERFEIT BANKNOTES USING

NEURAL NETWORK
E.HARIPRIYA1, Dr. K. ANUSDHA2
1
M.Tech, DEE, Pondicherry University, Puducherry, India.
2
Assistant Professor, DEE,Pondicherry University, Puducherry, India.

Abstract— Banknotes notes are issued by the Reserve Bank of India (RBI), in the year 1934. They
are currently issued in the denomination of Indian rupees ₹10, ₹20, ₹50, ₹100, ₹200, ₹500, ₹2000. In
the proposed work, detection of the counterfeit banknotes using multispectral images is implemented
by Neural Network technique. It will reduce the cost when neural networks are used. It shows good
performance and a low level of complexity. From the results, by using Pattern Recognition algorithm
is used to train and test the image at an advanced level to compute the recognition rate level. So that,
the performance metrics are increased when compared to the existing work.

Keywords: Banknotes, Reserve Bank of India (RBI), Neural Networks, Pattern Recognition.

I. INTRODUCTION
In India RBI, is the one and only bank which has full authority rights to issue banknotes. Counterfeit
money is imitation currency produced without the legal sanction of the government. Producing or
using counterfeit money is a form of fraud or forgery. A form of counterfeiting is the production of
documents by legitimate printers in response to fraudulent instructions.
Money/Coins are used as the medium of exchange for goods and services. Indian currency consists
of seven major denominations (₹10, ₹20, ₹50, ₹100, ₹200, ₹500, ₹2000), with each having different
features such as size, prominent color, identification mark [3].
To prevent and detect the circulation of counterfeit notes and its system must be developed. Currency
with the legal sanction of the government have some certain security features such as intaglio
printing, fluorescence, watermark, identification mark, security thread, see through register, micro
lettering, latent image and year of printing as seen in Fig. 1
Over the recent few years, as a result of the great technological advances in color printing,
duplicating, and scanning, fake money is spread in system largely and problems have become more
and more serious [4].

Fig. 1 Different Techniques for Detecting Fake Currency

DOI:10.23883/IJRTER.2018.4425.NB6DM 77
 International Journal of Recent Trends in Engineering & Research (IJRTER)
Volume 04, Issue 12; December- 2018 [ISSN: 2455-1457]

II. EXISTING WORK

In this paper, the fake currency detection using the image processing is performed. It is a process of
finding the forgery currency. After chosen, the image preprocessing method is applied. In pre-
processing, the image is smoothened and adjusted. Convert the image into gray scale level. After
conversion apply the image segmentation. Finally, compare the image into original or fake.

FLOW CHART:

Fig. 1 Block Diagram

ALGORITHM:
STEP 1: Image currency of ₹.100 will be acquired as an input image.
STEP 2: RGB image converted as gray scale image.
STEP 3: Edge detection will be performed. After detecting edges, the currency will be sharpen and
segmented.
STEP 4: After segmentation, the characteristics of the paper currency will be extracted.
STEP 5: The characteristics of final output image of security thread (black stripes) are compared
with the original pre-stored image in the system.
STEP 6: If it matches then the currency is original otherwise fake.

INPUT IMAGE:

Fig. 2 Input Image

OUTPUT IMAGE:
In feature extraction, the security thread can be obtained to check the authenticity of the image.

@IJRTER-2018, All Rights Reserved 78

 International Journal of Recent Trends in Engineering & Research (IJRTER)
Volume 04, Issue 12; December- 2018 [ISSN: 2455-1457]

Fig. 3 Output Image

III. PROPOSED WORK

FLOW CHART:

Fig. 4 Block diagram

ALGORITHM:
STEP 1: Image of paper currency Rs.100, 200 and 2000 images will be acquired as an input image.
STEP 2: In pre-processing, noise can be removed by the smoothening process.
STEP 3: In gray scale conversion, the image is converted from RGB to gray scale.
STEP 4: In feature extraction, RGB image can be converted as BW image, from that we can identify
the original Black stripes and Identification marks to find the currency is genuine or counterfeit.
STEP 5: It compare the images, if it matches then the currency is genuine. And it goes to the Neural
Network technique.

IV. METHODOLOGY
ACQUISITION IMAGE:
The first stage of input is an image acquisition stage. Input of Rs.100, Rs.200, Rs.2000 original and
fake images are given as an input.

@IJRTER-2018, All Rights Reserved 79

 International Journal of Recent Trends in Engineering & Research (IJRTER)
Volume 04, Issue 12; December- 2018 [ISSN: 2455-1457]

Fig. 5 Input Images

PRE-PROCESSING:
In pre-processing, it enhances the images and improves the datasets, also said to be Image
restoration. It involves the correction of distortion and noise. More filter operations are used to
remove the noise by smoothening process.

Fig. 6 Pre-processing Images

GRAY SCALE CONVERSION:

The image is converted to gray scale level because it carries only the intensity information which is
easy to process instead of processing three components R (Red), G (Green), B(Blue).

Fig. 7 Gray Scale Images

@IJRTER-2018, All Rights Reserved 80

 International Journal of Recent Trends in Engineering & Research (IJRTER)
Volume 04, Issue 12; December- 2018 [ISSN: 2455-1457]

FEATURE EXTRACTION:
It is a special form of dimensional reduction. Transforming the input data into the set of features is
called feature extraction. Finally the features are extracted through the security thread as black
stripes, fluorescence, see through the register, five angular bleed lines for ₹.200, seven angular bleed
lines for ₹.2000 and identification marks.

Fig. 8 Feature Extraction

NEURAL NETWORK
Neural networks is the ability to derive complicated data, can be used to extract patterns and
detect too complex to be noticed by either humans or other computer techniques. In this technique,
Pattern Recognition algorithm is used for training and testing the image at advanced level to compute
the recognition rate level.
Recognition Training Tested Data
Rate Data
₹ 100 100 % 91.66 %
₹ 200 100 % 91.66 %
₹ 2000 100 % 87.71 %
Table 1 Training and Tested Data Recognition rate

V. EXPERIMENTAL RESULTS:
Mean Squared Error Rate (MSE):
It quantifies the difference between an estimator and the true value of the quantity being estimated.
MSE = 1/MN [∑𝒏𝒊=𝟏(𝒚 − 𝒚𝟏) ^2] (1)

Root Mean Squared Error Rate (RMSE):

The square root of the MSE rate value is known as RMSE.
RMSE = sqrt (MSE) (2)

Peak Signal-to-Noise Ratio (PSNR):

The quality between the enhanced image and the original image.

PSNR(dB) = 10*log10 (256*256/MSE) (3)

@IJRTER-2018, All Rights Reserved 81

 International Journal of Recent Trends in Engineering & Research (IJRTER)
Volume 04, Issue 12; December- 2018 [ISSN: 2455-1457]

TABULATION:
Denominati Existing Value Proposed Value
on s s
Of Bank PSNR(d RMS PSNR(d RMS
Notes B) E B) E
100 31.70 6.65 42.51 1.91

200 39.17 2.81 41.08 2.26

2000 51.02 0.71 Infinity 0

Table 2 Comparison of Existing and Proposed work performances

VI. CONCLUSION
The proposed method is implemented by Pattern recognition in Neural Network technique. It reduces
the cost of fake banknote detection. This algorithm was trained and tested with banknotes. When
neural networks were used, the experimental results showed better accuracy for banknotes. It shows
good performance, low level of complexity, and more advantages that showed better recognition rate
level.

REFERENCES
[1] M.Deborah, C. Soniya Prathap “Detection of Fake currency using Image Processing”,International
Journal of Innovative Science, Engineering & Technology, Vol. 1 Issue10, December 2014.
[2] D.Alekhya, G.Devi Surya Prabha, G.Venkata Durga Rao, “Fake Currency Detection Using Image
Processing and Other Standard Methods”, International Journal For Technological Research In Engineering
Volume 1, Issue 11, July-2014.
[3] Sahana Murthy, JayantaKurumathur, B Roja Reddy, “Design and Implementation of Paper Currency
Recognition with Counterfeit Detection”, International Conference on Green Engineering and Technologies,
2015.
[4] Shital Mahajan, K.P.Rane, “A Survey on Counterfeit Paper Currency Recognition and Detection”,
International Journal of Engineering Research and Applications, April 2014.
[5] D.Alekhya, G.Devi Surya Prabha, G.VenkataDurgaRao, “Fake Currency Detection Using Image
Processing and Other Standard Methods”, International Journal For Technological Research In Engineering
Volume 1, Issue 11, July-2014.
[6] Allah BuxSargano, Muhammad Sarfraz, NuhmanUIHaq, “Robust Features And Paper Currency
Recognition System”, The 6th International Conference on Information Technology, 2013.

@IJRTER-2018, All Rights Reserved 82