Automatic number plate recognition system using modified stroke width transform

Abstract

Automatic number plate recognition(ANPR) techniques occupy a significant role in intelligent traffic management systems. Most of the existing ANPR systems do not perform well under conditions like high vehicle speed, varying illuminations, changing back ground etc. In this paper we propose an ANPR technique which efficiently solves the above issues using the stroke width transform. In our method, first we pre-process the image to remove blurring artifacts. Then the significant edges in the image are detected and a new image is formed by grouping the connecting rays. This is followed by morphological processing which detects the letters in the image. Then an appropriate Optical Character Recognizer (OCR) recovers the final result. Experimental results show that the proposed method work effectively under varying light conditions and even in the presence of shadows.