IMPLEMENTATION OF FAKE CURRENCY DETECTION USING

IMAGE PROCESSING AND SVM ALGORITHM

Dr . K.V. Venkata Kumar K.Mukhesh M.Sathwik

Dept. of Electr. & Commn. Engg Dept. of Electr. & Commn. Engg Dept. of Electr. & Commn. Engg
Vignan’s Lara Institute of Technology Vignan’s Lara Institute of Technology Vignan’s Lara Institute of Technology
& Sciences(VLITS) & Sciences(VLITS) & Sciences(VLITS)
Guntur,A.P.,India. Guntur,A.P.,India. Guntur,A.P.,India.
venkatkumarucet@gmail.com kalvapudimukesh@gmail.com msathwik9677@gmail.com

P.Sai Lakshmi K.Uday Sai Kiran

Dept. of Electr. & Commn. Engg Dept. of Electr. & Commn. Engg
Vignan’s Lara Institute of Technology Vignan’s Lara Institute of Technology
& Sciences(VLITS) & Sciences(VLITS)
Guntur,A.P.,India. Guntur,A.P.,India.
sailakshmipaidimarri@gmail.com udaysaikiran021@gmail.com

ABSTRACT: The advancement of color printing software, aims to determine the authenticity of
technology has made it much easier for banknotes, addressing this challenge.
counterfeiters to produce fake currency notes, a With rapid technological advancements in
problem that has escalated in India due to issues the financial sector, automatic cash identification
like black money and corruption. Previously, has become increasingly vital, particularly in
counterfeiting was challenging, but now, with vending and sales machines. There's a growing
laser printers, it has become accessible to almost advocacy for developing dependable cash
anyone. To address this, a system is being identification systems to meet this need. Nowadays,
developed to quickly detect fake currency notes it's commonplace to encounter machines capable of
using image processing techniques. This system recognizing banknotes in vending machines for
will analyze various features of Indian currency items like beverages, snacks, and tickets. The
notes, such as Bleed Line, Watermarking, primary objective of such systems is to extract
Fluorescence, Security Thread, Micro Lettering, relevant information from currency notes.
and Identification Mark. MATLAB software, While various methods exist for banknote
along with the Support Vector Machine (SVM) identification, image processing emerges as the most
model, will be utilized for feature extraction and effective. However, current methods may only
data analysis. The system's advantages include its achieve an accuracy of around 90%. To enhance
speed and simplicity, enabling it to accurately accuracy, the proposed system integrates the
determine the authenticity of currency notes. Support Vector Machine (SVM) Algorithm. This
Accuracy is a key metric in evaluating the system's improvement seeks to optimize the system's
performance. performance and reliability in discerning genuine
from counterfeit banknotes.
INTRODUCTION
METHODOLOGY
As the demand for banknote recognition
systems rises, researchers are striving to develop The block diagram of proposed system is as shown
reliable methods to automate the process while in a below fig.1.
maintaining accuracy and speed. These systems find Steps involved in the block diagram are explained
applications in various areas such as vending below:
machines and retail transactions, necessitating robust • Image Acquisition
recognition software. Extracting essential data from • Pre-Processing
currency notes poses a significant challenge for • RGB to Gray-Scale Conversion
system designers due to the complexity involved. The • Edge Detection
proposed solution, implementable through MATLAB • Image Segmentation
 • Feature Extraction main goal of edge detection is to pinpoint significant
1) Statistical Features transitions in image brightness, which allows for the
2) Edge Features capture of crucial details and changes in the visual
environment. Detecting edges is essential for
• Classification using SVM Algorithm
identifying important regions of interest (ROI)
within an image, enabling further analysis and
RGB to
Image Acquisition Pre-Processing Gray-Scale Edge Detection operations in subsequent stages.
Conversion Image segmentation:
Image segmentation is a fundamental
Image Segmentation Image Segmentation technique in digital image processing where an
image is divided into multiple segments or regions,
each comprising a set of pixels. This process is often
Statistical Features Edge Features
Extraction Extraction referred to as image thresholding, where a specific
threshold value is determined. If the pixel value
exceeds the threshold, it is converted to white;
Classification using otherwise, it is converted to black. This method
SVM Algorithm
helps distinguish different parts of the image based
on their intensity or color, enabling clearer
Fig.1.Block diagram of proposed system
identification and analysis of specific areas or
objects within the image.
Image Acquisition:
Feature extraction:
The system operates by taking an image of
Feature extraction involves transforming
the currency to be verified as authentic. This input
input images into a distinctive set of useful features.
image can be obtained through methods such as
Essentially, it's a way to condense raw image data
scanning or capturing a photo with a mobile device,
into a more manageable and meaningful format.
which is then uploaded to the system for processing.
Efficiently extracted features streamline the process
Pre-processing:
of classifying images through formal methods.
Pre-processing involves preparing the
However, identifying these useful features can be
image before further analysis. One common method
challenging and time-consuming. Different methods
is the Median Filter, which is great for reducing
are available for extracting features from data,
noise, especially 'salt-and-pepper' noise where
including techniques like local binary patterns,
random pixels appear as black or white spots. This
transform features, principal component analysis,
filter works by calculating the median value of the
decision boundary feature extraction, and statistical
surrounding pixels for each pixel in the image,
features. For currency image characterization, it's
effectively smoothing out irregularities.
crucial to carefully extract relevant image features
RGB to Gray-Scale Conversion:
that serve as input for image classification methods.
Converting a color image to grayscale
The following features are extracted here:
involves understanding how colors are represented
in the image. In a color image, each pixel is a blend
of three primary colors : Red, Green, and Blue
(RGB). On the other hand, a grayscale image
consists of a single layer representing the intensity
of light or darkness at each pixel. Various methods
exist for converting a colored image to grayscale.
Edge detection:
Edge detection is a fundamental method in
image processing that helps find the boundaries Fig.2. Security features of Indian Currency notes
between different objects within an image. It works Fake currency detection system will be
by identifying abrupt changes in brightness, which varied according to the banknotes features of
typically indicate the presence of an edge. This currency note of the particular country. For Indian
technique is widely used in tasks like segmenting currency notes the features are as follow.
images and extracting data, particularly in fields for For experimenting, purpose takes Rs.500 note For
example machine vision and computer vision. The specially important security signs are as follows
See through register: Identification Mark:
See through register shall be seen on the Each currency note is distinguished by a
currency note in denominational numeral at the left unique identification mark. These marks come in
down to the observer. various shapes, serving as identifiers for different
Latent image : denominations: a triangle for Rs. 100, an H-symbol
When you observe the note, you'll notice a for Rs. 200, a circle for Rs. 500, and a rectangle for
latent image situated on the left side of Mahatma Rs. 2000. Positioned to the right of the watermark,
Gandhi's portrait within the vertical band these identification marks facilitate easy recognition
corresponding to the denomination number. This of the denomination.
latent image becomes visible when the note is held Ashoka pillar emblem:
vertically at eye level. The Ashoka pillar emblem is positioned on
Devnagari : the right side at the bottom of the currency note.
Denominational numeral in Devnagari at Serial Number:
the left side of currency note Orientation is The number panel on the currency note
horizontally on left side of Mahatma Gandhi portrait displays numerals arranged in ascending order,
Bleed Line: starting from small at the top left side and gradually
The obverse side of certain Indian currency increasing in size towards the bottom right side.
notes, specifically the 2000, 500, 200, and 100 rupee 1)Statistical Features:
denominations, features a helpful aid for the visually Statistical features pertain to consistent
impaired known as the bleed line. This line is printed patterns or structures present within raw data. These
on both the upper left and right-hand edges of the features are typically extracted using diverse
notes. For the 500 rupee note, there are a total of five methodologies, encompassing structural, statistical,
bleed lines on both the left and right sides, all printed and transform-based techniques. In this research,
in raised print. statistical-based methods for feature extraction are
Mahatma Gandhi: utilized, emphasizing first-order histogram-based
The relative positioning and the orientation features and second-order co-occurrence matrix
of Mahatma Gandhi's portrait on the currency note features obtained from MR images. Histogram
should remain unchanged. analysis, a fundamental aspect of first-order
Security Thread: statistics, furnishes crucial insights into the
The security thread, a prominent feature on distribution of pixel intensities and their frequencies
the currency note, is located to the right of Mahatma within the images.
Gandhi's portrait. This thread displays visible 2)Edge Features:
elements such as "RBI" and "BHARAT". When the Given that features are typically small and
note is held up to light, the security thread becomes exhibit varying structures with dimensions smaller
apparent as a single, unbroken line. than their diameter, the following shape-based
Governor's signature: features are extracted to characterize them:
The guarantee clause, along with the • Area
Governor's signature featuring the Promise Clause • Convex area
and the RBI emblem, has been shifted towards the • Solidity
right on the currency note. • Extent
Water Mark: • Circularity
The watermark of Mahatma Gandhi is a • Ellipticity
notable feature on the banknotes. It appears with a
• Eccentricity
shaded effect and consists of multidirectional lines Classification using the SVM Algorithm:
within the watermark design.
The SVM algorithm involves the utilization of
Denomination in Green color:
supervised learning models and associated
The denomination in numerals,
algorithms for analyzing data in both classification
accompanied by the Rupee symbol, is printed using
and regression tasks. Given a set of training
color-changing ink, transitioning from green to blue. examples labeled with their respective categories,
This feature is located on the bottom right corner of
SVM constructs a model to categorize new
the banknote.
examples into one of the given categories. It
functions as a non-probabilistic binary linear
classifier, although methods like Platt scaling can RESULT IMAGE-3:
extend its application to probabilistic classification
settings.
In our system, we identify a concise set of features
for each category, including statistical and edge-
based features. Upon encountering a counterfeit
currency image, our system computes these specific
features for each category. These extracted features
are then fed into SVM classifiers to discern between
fake and genuine currency notes.

SIMULATION RESULTS

RESULT IMAGE-1: Fig.5.Gray Image

Gray Image: In the above fig.5 we convert the
denoised image into gray scale image. Pixel color is
typically represented through the combination of
Red, Green, and Blue (RGB) channels.
RESULT IMAGE-4:

Fig.3. Resized Input Image

Input Image: Taking a picture with the phone and
giving it as an input image. The given image is
resized and given as input image for processing.
RESULT IMAGE-2:

Fig.6. Edge Image

Edge Detection Image: Edge detection is an image
processing technique utilized to identify the
boundaries or edges of objects within an image.
RESULT IMAGE-5:

Fig.4. Denoised Image

Denoised Image: The above fig.4 shows the
denoised image. Denoised image is nothing
removing the noise from the given image for not to
degrade the quality of the image in further
processing steps. Here we use Median filter for
removing the noise. The application of a median
filter is commonly recognized as a robust method for
Fig.7. RIO Feature-Extraction(Texture)
noise reduction. It proves highly effective in
ROI FEATURE-EXRACTION(TEXTURE): In
mitigating 'salt-and-pepper' noise, demonstrating
image processing, a feature extraction program is
resilience against outliers in gray-level presence.
employed to detect and extract diverse visual
characteristics from an image, encompassing
shapes, colors, and edges among others form in input REFERENCES:
image.
[1] Swapnil,Gharat, Shamika Desai, Atharva
RESULT IMAGE-6: Rajadhyaksha, and Anjali Shetty. "CNN based
Counterfeit Indian Currency Recognition Using
Generative Adversarial Network." 2021.

[2] Bhagyavanitha Rajulapati, Srinivasulu Raju S,

Nikhil Vanteru, Sai Tarun Kara, and Sowmya Loya.
"Detection of Fake Indian Currency Using Deep
Convolutional Neural Network." 2023.

[3] T. Venkat Ram Reddy, Reddy B. Jayanth, and R.

Sumalatha. "Identification of Fake Indian Currency
using Convolutional Neural Network." 2022.

Fig.8. RIO Feature-Extraction(Edge & Shape)

[4] Neeru Rathee, Yatin Jaie, Rajat Sachdeva, Arun
ROI FEATURE-EXRACTION(EDGE &
Kadian, and Vijul Dalel. "Feature fusion for fake
SHAPE): This feature extraction program tasked
Indian currency detection." 2016.
with identifying the edges of an image would display
the detected edge details.
[5] Pranshu Mishra, Mandeep Kaur, Nitin Rakesh,
RESULT IMAGE-7: Parma Nand, Payal, and Pravidhi Garkoti. "Indian
Currency Recognition System Using Image
Processing Techniques." 2022.

[6] Ashok Kumar, Vivek Bhatnagar, Santosh

Kumar, Anshoo Bhatia, Savya Sachi, and Pravesh
Belwal. "Fake Currency Detection using Ensemble
Learning." 2023.

[7] Bhupesh Kumar Singh, Kuldeep Verma, and

Anupam Agarwal. "Indian currency recognition
based on texture analysis." 2011.
Fig.9. Currency Type : Real
Real Note: If the currency note is real then Currency [8] Vedasamhitha Abburu, Saumya Gupta,
Type : Real is displayed in dialogue box S.R.Rimitha, Shashidhar, G. Koolagudi, and
Manjunath Mulimani. "Currency recognition system
RESULT IMAGE-8: using image processing." 2017.

[9] G.R. Gidveer and Sonali R. Darade. "Automatic

recognition of fake Indian currency note." 2016.

[10] Elham Dallalzadeh, Amir Rajaei, and

Mohammad Imran. "Feature Extraction of Currency
Notes: An Approach Based on Wavelet Transform."
2012.

[11] Kalpna Gauta. "Indian Currency Detection

using Image Recognition Technique." 2020.
Fig.10. Currency Type : Fake
Fake Note: If the currency note is fake then
Currency Type : Fake is displayed in dialogue box.