BOUNDING BOX METHOD BASED ACCURATE VEHICLE NUMBER DETECTION AND RECOGNITION FOR HIGH SPEED APPLICATIONS

Abstract

License plate detection and recognition is the one of the major aspects of applying the image processing techniques towards intelligent transport systems. Detecting the exact location of the license plate from the vehicle image at very high speed is the one of the most crucial step for vehicle plate detection systems. This paper proposes an algorithm to detect license plate region and edge processing both vertically and horizontally to improve the performance of the systems for high speed applications. Throughout the detection and recognition the original images are detected, filtered both vertically and horizontally, and threshold based on bounding box method. The whole system was tested on more than twenty five cars with various license plates in Indian style at different weather conditions. The overall accuracy rate of success recognition is 93% at sunlight conditions, 72% at cloudy, 71% at shaded weather conditions.