An Efficient Approach for Automatic Number Plate Recognition for Low Resolution Images

Abstract

Apart from the developed countries, many developing countries are facing the wrath of internal and external security. Inadequate resources force these countries to adopt the usage of Automatic Number Plate Recognition (ANPR) system to ramp up security to a certain extent. These countries use CCTV cameras for manual surveillance. The proposed ANPR system is designed to work with these CCTV cameras. Fundamental components of the image are for image enhancement, character segmentation and character recognition. Novel techniques for each of these components are the main contributions of this work. The proposed approach has been tested on both low and high resolution images. In low resolution images, the average accuracy in character recognition achieved by the proposed approach is 96.2% and the average accuracy in character segmentation is 100% whereas the other existing techniques are unable to recognize the number plates. In good quality images, the proposed approach achieves 100% accuracy in character segmentation and 95.3% average accuracy in character recognition. The approach achieves 4.5% improvement over other existing approaches in character segmentation for good quality images.