Malaysian License Plate Recognition Algorithm Using Convolutional Neural Network

• Main Author: Piramli, Muhamad Marzuki
• Format: Thesis
• Language: English; English
• Published: 2020
• Subjects: T Technology (General); TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://eprints.utem.edu.my/id/eprint/25422/1/Malaysian%20license%20plate%20recognition%20algorithm%20using%20convolutional%20neural%20network.pdf; http://eprints.utem.edu.my/id/eprint/25422/2/Malaysian%20license%20plate%20recognition%20algorithm%20using%20convolutional%20neural%20network.pdf; http://eprints.utem.edu.my/id/eprint/25422/; https://plh.utem.edu.my/cgi-bin/koha/opac-detail.pl?biblionumber=119752

Nowadays, License Plate Recognition (LPR) becomes popular among researchers due to its compatibility in many applications. For instance, LPR significantly can be applied to toll gate systems, surveillance systems and law enforcement. However, the previous LPR system still does not meet optimum accuracy and speed. Current Convolutional Neural Network (CNN) improvements have the ability to solve complex visual recognition tasks. The primary aim of this system is to ensure that the character of the vehicle plate recognize accurately and efficiently using CNN techniques. A method utilizing two CNN network architectures of deep object detection was designed to solve the Malaysian License Plate Recognition (MLPR) task. The first and the second network were designed for plate detection and recognition of license plate characters respectively. Both of the networks utilized the architecture of YOLOv2 with high speed and accuracy. The accuracy and speed of the plate recognition of the MLPR obtained were 98.75% and 0.0104 seconds respectively. The MLPR has obtained high prediction accuracy and has outperformed the existing methods. In conclusion, the system adapted from deep object detection is the best solution for the MLPR problem based on the accuracy and speed achieved.