Modified Coin Identification Using Neural Network

In recent years there has been a growing trend of using neural networks for the development of intelligent systems that are able to simulate pattern recognition and object identification. Physical parameters of a coin are the sole criterion that is currently being used for Coin identification by machines. As physical features can easily be easily imitated another criterion which is not based on physical features but recognises the pattern of the coin helps preventing confusion between different coins of similar physical dimensions and against forgery. The paper purposes a rotation-invariant coin identification system (RICIS) that adds the features of the physical properties like colour, radius and centre to recognize rotated coins by 5 degrees and at the same time advantages of pattern matching using a neural network and pattern averaging along with the physical features. This Rotation Invariant Coin Identification System, abbreviated as RICIS, firstly preprocesses the image, where meaningful representations of coin patterns are provided removing the background and the unnecessary data within the result images. In the second phase the coin pattern is learned when the optimized data is fed to a resilient back propagation neural network representing the coin images. RICIS has been successfully implemented as shown in this work to identify the 1, 2,5,10 rupee coins. This solves a real life problem where physical similarities between these coins led to slot machine abuse. An overall 99.4% correct identification of coins has been achieved. Keywords— Digital Signal Processing, Feature Vector Table, Image Processing, Neural Network