Vehicle License Plate Recognition System

Abstract

This paper presents a method for automatic detec tion and recognition of vehicle license plates in a Sri Lankan context. License Plates of the still images capturing the frontal view of vehicles are initially located by identifying the regions with the highest vertical edge density. After applying dilation and connected component analysis, the license plate can be localized. The Hough Transformation is used on the extracted plate region to find the skew angle which is used to rotate the plate region to align it with the horizontal axis. Then, the intensity of the plate region is enhanced using morphological operations before obtaining the corresponding binary image. The plate area is further pre-processed before extracting the characters. Boundaries of the characters are identified by the row sum of pixels and column sum of pixels. Extracted characters are normalized to 16 x 10. Then, template matching is used to recognize the characters of the license plate. The developed license plate recognition system has an accuracy of 84% for detecting the number plates of the still images of vehicles.