Automatic Bengali license plate detection and recognition using neural networks

Abstract

Automatic license plate detection and recognition has become one of the obligatory components in the field of smart traffic control systems in Bangladesh. However, Automatic License Plate Recognition (ALPR) can also be bene cial for parking lot management systems, detecting stolen vehicles, toll management systems, nding convicted vehicles, which are involved in road related violations etc., and other purposes. In order to ease the smart tra c control system, the license plate detection and recognition process needs to be very e cient.Various methods have already been introduced in order to make the detection and recognition process e cient. Since there have been very few works conducted on Bangla license plates, these methods are not quite e cient due to wide variation in Bangla license plates. Therefore, we have developed a new method using multiple algorithms to increase the e - ciency of detection and recognition process. Our proposed method is composed of three stages which are Detection, Segmentation and Recognition likewise most of the conventional license plate recognition systems. We have applied di erent ex- isting algorithms such as Canny edge detection algorithm, Otsu image binarization algorithm etc. in di erent stages. We have achieved an accuracy of around 95% by using our proposed method.