Reliable License Plate Recognition
Abstract
This paper presents a new method for the detection and recognition of the Persian-Arabic vehicle license plate system from outdoor images. The methods designed with multithreading architecture for implementation on a parallel architecture. The basis system uses symmetry for vehicle detection, line sweeping for plate detection and fuzzy method for plate recognition. The main theme is to use a novel Multilevel Illumination-Dim Model for image evaluation in such a way that the plate and its number can be extracted accurately with an inference Reliability Rate of Recognition. The method makes the extraction of the plate independent of color, size, and the location of license plate in a vehicle image. The camera didn’t focus in the plate so the system is insensitive to camera angle of view and camera-vehicle relative position within a reasonable distance. The experimental results on 500 vehicle color images reached 98.68% for plate recognition, 100% for plate detection and 100% for in existence of vehicle or plate in the picture. Weather and lighting conditions, normal imaging distance, reasonable range of illumination, and angle of view didn’t affect efficiency. Better performance with improvement in speed and accuracy has been reported while limitations in distance, angle of view, illumination conditions are set and background complexity is low.
Keywords
License Plate Recognition (VLPR), Reliability Rate of Recognition (3R), Persian Optical Character Recognition (POCR), Multilevel Illumination-Dim Model (MIDM), Intelligent Transportation Systems
