Development of a Nigeria Vehicle License Plate Detection System

Abstract

The importance of license plate detection system cannot be overemphasized in intelligent transport systems. License plate is a major component in most of the applications related to intelligent transport system. Moreover, it is also a quite popular and active research topic in the field of computer vision and image processing. Different techniques and algorithms have been proposed to detect license plate number from a vehicle image. Nevertheless, due to the variation in climate conditions, characteristics of the license plate, numbering system, colors, fonts and size, further work is still needed in this field in order to make the detection and recognition process accurate and very efficient. For these reasons, this paper presents a scheme for license plate detection using current image processing techniques. The developed scheme used images obtained from Caltech database and our newly acquired Ahmadu Bello University (ABU) dataset. To detect the license plate, the acquired images were pre-processed to reduce the computational requirement of the developed scheme. Canny operation is performed to detect the edge of the pre-processed images then histogram equalization is applied to spread out the contrast of the image. Edged information is used to extract the region which constituted the license plate number and lastly Support Vector Machine is used to distinguish the true license plate from other regions. The performance of the developed scheme is evaluated on the Caltech dataset and the ABU dataset. The experimental result shows that our model achieved a better detection rate accuracy than some existing methods.