An Approach for Automatic Recognition System for Indian Vehicles Numbers Using K-Nearest Neighbours and Decision Tree Classifier^

Abstract

An Automatic Vehicle Number Plate Recognition System is presented in this paper. Basically, there are different kinds of motor vehicle number plates which are used in India. The actual number plates are not similar and most of them are deviated from the prescribed norms as laid by Government of India from time to time. The number plates of Indian vehicles are written in different languages also. The number plate characters are varied according to the language of a particular state in the country however a large portion of number plates are written in English language (legally all plates are required to be in modern Hindu-Arabic numerals with Latin letters). This is very vital in recognizing the number plates, extraction of number plate characters, segmenting the characters because of some diversity. In this paper we present a work on extraction of number plates, segmentation and recognition of number plates in the domain of English language. Prewitt filter technology is used for extracting the number plate thresholding and for segmentation of connected component analysis are used. For recognizing the characters, neural network, k-nearest neighbor and decision tree classifiers are used. The recognition rate of neural network is 96.1%, k-nearest neighbor is 95.55% and decision tree is 91.38% respectively which is acceptable for future research in this direction.