Currency Recognition System

A software for visually impaired that helps them recognize various Indian currency denominations of Rs. 50, Rs. 100, Rs. 500 and Rs. 1000

Image Processing
Image Processing

Motivation


The software helps classify the input currency denomination into Rs. 50, Rs. 100, Rs. 500 and Rs. 1000. The motivation behind the software was to finally embed it in a hardware prototype so that visually impaired community could use this tool to differentiate between various denominations of currency to carry out their day to dat exchange of money. I was the sole developer for this piece of software.

Technology


Bag Of Words algorithm is used to implement this software; SIFT descriptor is used to retain the key point features. Bag of Words algorithm helps in plotting a histogram of occurrence of keypoints. These histograms are then send to 1-v/s-all SVM classifier. The vocabulary trained is 1000 feature-long. The classifier is trained on 1000 images and tested on 700 images. The average accuracy of the module is 88.25%.

Result


Accuracy for Different Currency Denominations
Currency (Rs.)
Accuracy
50
85.06 %
100
100 %
500
77.95 %
1000
90 %