An Improved Design for Detection of Malaysian License Car Plate Number

Abstract

Vehicle license plate identification system is an image processing technology used to identify vehicles by their license plates. The key step in vehicle license plate identification system is plate region localization which is a process of finding the region of interest (plate number) from an image. As a prior stage before recognition, localization is considered the crucial stage in any license plate identification/recognition system whose overall performance of the system depends on it. License car plate standardization varies from one country to another. Some of the Malaysians’ car plates exhibit non-conformity with the standard in production which made them difficult to be recognized, as such, they remain a threat to the performance of many identification systems designed so far. Against this background, this project aims towards coming up with an improved design flow for the localization of vehicle plate right from the detection stage using Difference of Gaussian operator (D-o-G) for edge detection, skew correction using Radon matrix and image binarization for identifying the true plate number region. For efficient and better-desired result delivery, this project was designed to take care of only the detection phase of the identification system. At the end of the implementation, a promising result was obtained over the detection of many of the plate with correct detection of double row plates which failed in the previous design. As an improved design of the previously adopted design, the performance accuracy achieved was 90.25% with few failures on some single row plates.