Date of Award

1998

Degree Type

Thesis

Degree Name

Master of Engineering

School

School of Engineering

Faculty

Faculty of Science, Technology and Engineering

First Advisor

Assc. Professor Clifton Smith

Second Advisor

Dr Stefan Lachowicz

Abstract

The development of automatic car registration plate recognition systems will provide greater efficiency for vehicle monitoring in automatic access control, and will avoid the need to equip vehicles with special RF tags for identification since all vehicles possess a unique registration plate. Thus this study is an attempt to introduce an automatic car registration plate recognition system based on identifying the plate characters by using the Hough transform. However, the proposed recognition system can be used in conjunction with a tag system for higher security access control. The automatic registration plate recognition could also have considerable potential in a wide range of applications especially in the identification of vehicle-based offences and with law enforcement. Recent advances in computer vision technology and the falling price of the related devices has contributed in making it practical to build an automatic, registration plate recognition systems. There have been a number of Optical Character Recognition (OCR) techniques, which have been used in the recognition of car registration plate characters. These systems include the character details matching process (Lotufo, et al. 1990), BAM (Bi-directional Associative Memories) neural network (Fahmy 1994) neural network (Tindall, 1995) and cross correlation pattern matching character matching techniques (Cornelli, et al. 1995). All of these systems recognized the characters by matching the full image of every character with a character's template database which requires considerable processing time and large memory for the database. The purpose of this study is to explore the potential for using Hough transform (Hough 1962) in vehicle registration plate recognition. The OCR technique used in this project is unlike the other systems where the character recognition was based on matching the character's full image; However the OCR technique in this system used Hough transform to identify the characters, where the recognition of a character is based on matching its identification array to the database. To validate the research, a car registration plate recognition system was developed to locate the registration plate from the full image of a vehicle and then extrar.t the plate characters by using image processing techniques. A Hough transform algorithm was applied to every character within the registration plate image to produce an identification array for these characters, and the plate characters were recognized by matching their identification array to the database. The system has been applied to a number of video recorded car images to recognize their registration plates. The rate of correctly recognized characters was 82.7% of the extracted characters, but improvement can be granted by using a faster digital camera and taking some precautions in the registration plate frames. However, the research indicated that the optical character recognition technique used in the study is an efficient and simple algorithm to identify characters, without requiring a relatively large processing memory.

Share

 
COinS