Recognition System for Pakistani Paper Currency

: There are many real-life applications which heavily use many techniques based on Pattern Recognition such as voice recognition, character recognition, handwriting recognition and face recognition. Paper currency recognition is a new application of pattern recognition. This application uses the computing power in differentiating between different kinds of currencies with their suitable class. Selection of proper feature enhanced the performance of the overall system. We are aiming to develop an intelligent system for Pakistani paper currency that could recognize the currency note accurately. In this paper, we have taken samples domain of five different Pakistani paper currency notes (Rs. 10, 20, 50, 100, 1000). We scanned total 100 currency notes, 20 from each sample of selected domain for feature extraction of these images using a software. The images will be matched with the features stored in MAT file and if the features of test images will be matched with that file, the software will return the class of that currency note. Experimental results are presented which show that this scheme can recognize currently available 8 notes of Pakistan’s Currency (Rs. 10, 20, 50, 100, 500, 1000 etc.) successfully with an average accuracy of 98.57%.


INTRODUCTION
the color and pattern.The staffs who work for the money We live in an age of information where everyone is an easy job.This may cause some problems (e.g.Wrong busy and life becoming faster day-by-day.In this busy recognition), so they need an efficient and the exact life everyone needs complete and quick and correct system to help their work.As we mentioned before, the response so they can save their time.aim of our system is to help people who need to recognize Today's era is termed as the "IT age" or we can say Pakistani paper currency and work with convenience and it "Computer's Age" so it is thought inevitable to have efficiency.software solutions for various problems in order to save Otherwise our system is based on image processing, the time, in different organizations despite of it whether techniques which include Noise removal, preprocessing, the organization is small or large enough.We focus to In order to make the system more comprehensive, we need recognize Pakistani paper currency accurately.
to create a small database for storing the features of the Paper currency recognition is one of the applications currency.The system will be programmed based on of pattern recognition.There are some similar recognition MATLAB and include a user-friendly interface.The main systems, such as face recognition system, fingerprint steps in the system are: recognition system.However the theories they use are similar but the techniques and approaches are Read image, reading the image we get from the different.
scanner as well as the format of the image is JPEG.The Pakistan paper currency has different Preprocessing, removing noise.denominations, with each of them looking totally different.
Feature extraction, classification.For instance the size of the paper is different, the same as Result.
have to distinguish different type's notes and that is not World Appl. Sci. J., 28 (12): 2069-2075, 2013 2070 There are 7 denominations of Pakistani paper currency is divided into n equal parts along vertical currency.Each note has different size and different color.
vectors for these parts.This system is designed to reduce the human effort and In [4] they have represented a currency recognition to avoid the purchase of expensive hardware.This system system using negatively correlated ensemble neural will extract the features of the test image and will match network.They have proposed the ENN for currency with the features stored in training database (mat file).
recognition.For training they used the negative If the features match it will display the type of currency.correlation learning.In negative correlation the entire There is no as such system for recognizing Pakistani networks are negatively correlated through the strength Paper Currency.This system can be used in: of penalty term.The entire ensembles interact with each ATM Machines portion of input vector.So when they apply a noisy Auto-Seller Machines pattern the network will be able to recognize as a whole.Bank Money-Counters.
The recognition rate is better than single network.
The scope of the project is to recognize the Pakistani Bangladeshi currency but it is equally applicable in any paper currency note correctly and accurately.The main paper currency recognition.objectives of this is to develop an intelligent system for In to technical limits and computational memory.Heuristic The neural network can recognize patterns effectively analysis of characters of the serial number is done.It is a and robustly.In this they use a new kind of banknote technique which actually produces a good solution.Thai bank note as the object of recognition.In their anti-virus scanners use heuristic signatures to look for recognition system masking process is defined as the specific attributes and characteristics to detect virus.characteristics extraction of a bank note image.Neural By using heuristics, time can be reduced when network learning and recognition algorithms are solving problems .As because the heuristics are fallible, implemented On DSP devices as a Neuro recognition it is important to understand their limitations.They are engine; they proposed the continuous learning by the used as acids to make quick estimates and preliminary DSP unit which they have developed for banking process designs.machines.
In [6] they considered the distinctive point extraction In [2] they have worked for recognition of various and recognition algorithm for various kinds of banknotes.kinds of paper currency.They proposed multiple kinds of By converting the scanned 256-colored image data to 4-bit paper currency with neural network using axis symmetrical gray data as pre-processing, we can get a better algorithm mask and two sensors.Mask extracts the characteristics to find the dark areas on the special block because the of paper currency they used two sensors to get both dark color is robust to noise.By applying the continuous images of currency surface.
same colored area recognition algorithm to the face value They construct an experimental system using a bank of the banknote, we can extract distinctive data to classify machine whose sensor is attached up side and down side.
the kind of banknotes, as the area is located in the Finally they applied the proposed method to Euro different positions on each kind of banknotes.To currency.
recognize banknotes, we trained 5 neural networks.In [3] they use symmetrical mask for recognizing One is for inserting direction and the others are for the paper currency.In this method non masked pixel value of face value the distinctive data pattern according to the banknotes is computed and feed to the neural network for inserting direction shows relatively clearer tendency than recognizing paper currency.For this two sensors are used that of the face value.With this method, we can get a high at the front and back of paper currency but decision is recognition rate except for 100 and 200 Euro bank notes.done by the image of the front.In next step paper The proposed recognition algorithm does not include other and each network has specialized for a particular Though they have performed our experiment for information from the raw data which is relevant to classify position correction.In banknote counting machines, the origin position of the distinctive points may be changed when banknotes are not perfectly inserted into the counting machine.This occurs frequently and thus more researches will be needed.
The system is based on scanner, PC and classifiers.Here the classifier that we have used is Knn classifier that is in fact a nearest neighborhood classifier.Before applying a classifier we had used preprocessing techniques, in such a way that we had first converted the RGB image to grey scale image, in order to remove noise we had used wiener filter.The Wiener filter purpose is to reduce the amount of noise present in a signal by comparison with an estimation of the desired noiseless Fig. 3.1: Block diagram signal.Then the image is converted into a binary image to extract features.This system is designed under the System Architecture programming tool of MATLAB.
Acquisition of Images: We had scanned 100 Pakistani The software will extract the features of the test paper currency notes of different dominations.Each note images.Once these five features will be extracted they was scanned at 200 dpi using an HP scanner.will be matched with the features stored in MAT file.The features in MAT file are features of train images.
Preprocessing: The procedure is done before processing If the features of test image will be matched with the by correcting image from different errors, is features in MAT file the software will return the class of preprocessing.In this project the preprocessing that currency note.If the test image features don't match techniques used are conversion of RGB image to Gray, with any of the features in Mat file the software will removal of Noise and conversion to Binary.display that it is doesn't belong to any class.
In other words it will classify the paper currency to Conversion to Gray: It represents an image as a matrix the correct class to which it belongs (e.g., 10, 20, 50, 100, where every element has a value corresponding to how 500 and 1000).
bright/dark the pixel at the corresponding position The classification is the process to classify the should be colored.It assigns value between 0 to 255 to currency note into its correct class.For this purpose there represent the brightness of the pixel.0 represents black are many algorithms like Euclidean distance classifier, and 255 represent white.We convert RGB color image into Weighted Euclidean distance classifier , knn classify etc.
the gray level as color information is not useful in this the algorithm which we had used for classification is knn recognition process furthermore it reduces the classify.Instance-based classifiers such as the knn computational cost.classifier operate on the premises that classification of unknown instances can be done by relating the unknown Noise Removing: Many currency notes come with dust to the known according to some distance/similarity on them or something written on them.Noise removing is function.Classification (generalization) using an the process of removing such dust from these notes and instance-based classifier can be a simple matter of makes the image clearer.MATLAB includes many noises locating the nearest neighbor in instance space and removing filters.The filter which we used is wiener2.labeling the unknown instance with the same class label The Wiener filter purpose is to reduce the amount of as that of the located (known) neighbor.This approach is noise present in a signal by comparison with an often referred to as a nearest neighbor classifier.
estimation of the desired noiseless signal.Width: Width is the measurement of horizontal distance of an image.
MAT File: A MAT file is a Microsoft Access file.The purpose of using the MAT file here is to save the features of training images which we had extracted.The features will be used during classification.The features of test image will be matched with the features stored in MAT file.If features of the test image match with the features in MAT file the currency type will be displayed.

Classification:
The classification is the process to classify the currency note into its correct class.For this purpose there are many algorithms like Euclidean distance classifier, Weighted Euclidean distance classifier, knn classify etc. the algorithm which we had used for classification is knn classify.Instance-based classifiers such as the kNN classifier operate on the premises that classification of unknown instances can be done by relating the unknown to the known according to some distance/similarity function.Classification (generalization) using an instance-based classifier can be a simple matter of locating the nearest neighbor in instance space and labeling the unknown instance with the same class label as that of the located (known) neighbor.This approach is often referred to as a nearest neighbor classifier.
The software will extract the features of the test Fig. 4.1: Recognition of 10 rupee note images.Once these five features will be extracted they will be matched with the features stored in MAT file.The features in MAT file are features of train images.If the features of test image will be matched with the features in MAT file the software will return the class of that currency note.If the test image features don't match with any of the features in Mat file the software will display that it is doesn't belong to any class.

RESULT
Snap Shots 10 Rupee Note: First 10 rupee note test image will be given for recognition to the software.The features of that image will be extracted and will be matched with the features in mat file.If the features matched the software will recognize that it is the 10 rupee note and will display the type and the image.10different images of 10 rupee notes were given to this software for recognition and all of them were recognized successfully.Fig. 4.2: Recognition of 20 rupee note 20 Rupee Note: First 20 rupee note test image will be 50 Rupee Note: First 50 rupee note test image will be given for recognition to the software.The features of that given for recognition to the software.The features of that image will be extracted and will be matched with the image will be extracted and will be matched with the features in mat file.If the features matched the software features in mat file.If the features matched the software will recognize that it is the 20 rupee note and will display will recognize that it is the 50 rupee note and will display the type and the image.10different images of 20 rupee the type and the image.10different images of 50 rupee note were given to this software for recognition and all of note were given to this software for recognition and all of them were recognized successfully.
them were recognized successfully.100 Rupee Note: First 100 rupee note test image will be 500 Rupee Note: First 500 rupee note test image will be given for recognition to the software.The features of that given for recognition to the software.The features of that image will be extracted and will be matched with the image will be extracted and will be matched with the features in mat file.If the features matched the software features in mat file.If the features matched the software will recognize that it is the 100 rupee note and will display will recognize that it is the 500 rupee note and will display the type and the image.10different images of 100 rupee the type and the image.10different images of 500 rupee note were given to this software for recognition and all of note were given to this software for recognition and all of them were recognized successfully.
them were recognized successfully.1000 Rupee Note: First 1000 rupee note test image will be REFERENCES given for recognition to the software.The features of that image will be extracted and will be matched with the 1.Fumiaki Takeda, Lolita Sakoobunthu and Hironobu features in mat file.If the features matched the software Satou, 0000."Thai bank note recognition using will recognize that it is the 1000 rupee note and will neural network and continues learning by DSP unit" display the type and the image.10different images of 2. Fumiaki Takeda and Toshihiro Nishikage, 2000.1000 rupee note were given to this software for Multiple kinds of paper currency recognition using recognition and all of them were recognized successfully.
neural network and Applicable for Euro Currency.

Reliable Method for Paper Currency Recognition
This thesis shows the method for currency 4. Kalyan Kumar Debnath, Sultan Uddin Ahmed and recognition using image processing.The proposed M.D. Shahjahan, 0000."A Paper Currency system uses the different features of the currency for Recognition System Using Negatively Correlated recognition.Our experiment shows that this is the low Neural Network Ensemble".cost machine to recognize the Pakistani paper currency 5. Parminder Singh Reel, Gopal Krishan and Smarti notes.We had checked different notes on this system and Kotwal, 0000." Image Processing based Heuristic the result is 100% which means that the system is working Analysis for Enhanced Currency Recognition" efficiently.
6. Jae-Kang Lee and Hwan Kim, 0000." New Future Work: The future work will be done by applying Banknotes".different filters.In this thesis the images were scanned horizontally in the future the images will be scanned with different angles.Different currencies could be used for recognition like Indian Rupee, US dollar, EURO etc. Similarly different features can be used for recognition.
[5] the feature extraction of Indian currency notes Pakistani paper currency that could recognize the involves the extraction of features of serial numbers of currency note accurately.currency notes.Feature extraction means to extract the Related Work: In [1], they did recognition by using to minimize the class pattern variability and enhancing the neural network.The neural network is widely used between class pattern variability during feature extraction, for pattern recognition because of its abilities of the dimensionality of data is reduced and it is needed due self-organization, parallel processing and generalization.

Table 3 .
1: Number of Notes Scanned