AUTOMATED CAR PLATE LOCALIZATION FOR MOVING CAMERA
Authors: Ang Li Jun and Chin Kim On
In year 1976, license plate recognition system (LPR) was invented at the Police Scientific Development Branch in the United Kingdom. Prototype systems were deployed on the A1 road at the Dartford Tunnel. However, this system is not widely used due to the higher cost until year 1990s. Many researchers started to conduct researches that are related to LPR and many methods have been proposed. However, most of the researches conducted were focused on using fixed camera instead of using moving camera. As a result, it indirectly creates an initiative in me to carry out this research that focused on localize car plate in video frames capture with moving camera.
This research aims to (1) plan, design and implement a framework for video manipulation using MATLAB. A simple but easy to use interface is required in order to read selected video file, convert the video file into frames for further enhancement purpose, (2) plan, design and implement car plate localization techniques/algorithms to localize multiple car plates from the extracted frames by using edge based method and morphology based method, and (3) compare and analyse the proposed localization algorithms performances. There are two localization methods proposed in this research namely, edge based method and morphology based method. 250 frames with 333 car plates from 21 road scene videos with resolution format 1280*720 are used in experiment. The localization rates for both proposed methods correspond to 93.39% for edge based method and 90.99% for morphology based method.