License plate detection and character recognition system for commercial vehicles based on morphological approach and template matching

Abstract

Though license plate detection and recognition system has a lot of applications but a very little work has been done on Bangla license plate recognition. Variations among the license plate patterns and complex background of Bangladeshi license plates make it more difficult to use the existing algorithms. Because of this, we propose a solution for Bangla license plate detection and recognition. Firstly, the position of the vehicle is determined. In our country, commercial License Plates have the unique color (green) of its own. That's why; we select the portions of green color with the matching RGB intensity of the plate. A boundary based contour algorithm, area and aspect ratio have been proposed to track down the license plate in the vehicle region. License plate rows containing registration information have been separated by using horizontal projection with threshold. The characters and the digits of the rows have been segmented using vertical projection with threshold value. Finally, template matching has been used for recognizing the characters and the digits of the Bangla license plate. We tested our algorithms for over 180 still images captured from the road. We achieve93% success in license plate detection, 98.1% success in segmentation and 88.8% success rate in Bangla license plate recognition.