Automatic Number Plate Recognition System for Oman

Abstract

An automatic number plate recognition (ANPR) system is a key component of intelligent transportation system. It helps to manage huge number of mobile vehicles in roads, monitoring highways, parking management, and so other transportation applications. ANPR systems have the capability to automatically identify vehicle number plates from images and convert them to machine-readable ASCII characters. Systems have been developed using several algorithms and methodologies, including optical character recognition, convolutional neural network or deep neural network, morphological operations, and edge detection. This chapter discusses a proposed automatic recognition system for Oman’s number plates. For this purpose, this chapter presents a theoretical and analytical comparison between several previous works in this field to understand which algorithms are most suitable. A practical evaluation is conducted using actual number plates. According to this study, there are three main different levels of the recognition system number plate detection, number plate recognition, and character recognition. In each processing level, there are another subpreprocessing operations and deep learning algorithms, for example, using morphological operations on number plate detection, using thresholding operations to extract binary images in the level of number plate recognition, and using convolutional neural network in character recognition level. The performance of recognition operation is evaluated based on different metrics such as classification accuracy, logarithmic loss, F1 score, precision, and recall. Altogether, number plate extraction from vehicle images attained 71.5% accuracy, while character recognition based on extracted characters achieved 96–99% accuracy (depending on the type of character).