Automatic Number Plate Recognition using Random Forest Classifier

TL;DR

This paper presents an automatic number plate recognition system that uses image processing and a random forest classifier to achieve over 90% accuracy in identifying vehicle license plates.

Contribution

It introduces a novel combination of image processing steps with a random forest classifier for effective number plate recognition.

Findings

Achieved 90.9% accuracy in recognition.

Effective in noisy and low illumination conditions.

System processes images through four key steps.

Abstract

Automatic Number Plate Recognition System (ANPRS) is a mass surveillance embedded system that recognizes the number plate of the vehicle. This system is generally used for traffic management applications. It should be very efficient in detecting the number plate in noisy as well as in low illumination and also within required time frame. This paper proposes a number plate recognition method by processing vehicle's rear or front image. After image is captured, processing is divided into four steps which are Pre-Processing, Number plate localization, Character segmentation and Character recognition. Pre-Processing enhances the image for further processing, number plate localization extracts the number plate region from the image, character segmentation separates the individual characters from the extracted number plate and character recognition identifies the optical characters by using random forest classification algorithm. Experimental results reveal that the accuracy of this method is 90.9%.