Bangladeshi Vehicle License Plate Detection method based on HSI color model and geometrical properties

Abstract

Bangladeshi Vehicle License Plate Detection (BVLPD) plays an important and inevitable role in Vehicle License Plate Recognition (VLPR) system. The most challenging part of this method is to detect the region of the license plate from the vehicle image. In this paper, we propose an algorithm for analyzing the vehicle image to extract the LP position in the image of the vehicle. Initially, HSI color model is adopted to select a threshold for detecting candidate regions and then different geometrical properties of LP such as area, bounding box, aspect ratio are used to determine whether the candidate regions contain LP or not. Finally the candidate region is verified by intensity histogram. The proposed method is able to deal with candidate regions under different scale of the plate. In the experiment more than 100 images are used which are taken under different conditions such as uneven illumination, complex scenes, varied weather and varied distances from the vehicle to camera. The overall rate of success of the license plate detection algorithm is 85%.