Automatic Number Plate Recognition Technologies for Azerbaijani Vehicle Registration Plates

Abstract

Vehicle monitoring on road surveillance and parking lot management is one of the global problems in the computer vision field that requires the application of practical and contemporary technologies. The main applications in the computer vision field concerning these problems are called Automatic Number Plate Recognition Systems which are based on ever changing methodologies given the rapid development in Artificial Intelligence and Deep Learning fields. The reason for the broad spectrum of different methodologies is that these systems are in high demand by institutions and private businesses where ANPR systems are frequently upgraded to match the best practices available in order to improve performance, increase speed, reduce cost, and overall maximize the efficiency. Thus, this paper is focused on current practices in Automatic Number Plate Recognition systems and the performance of individual building blocks in the ANPR pipeline. In this paper we will examine in depth the details of contemporary practices in the computer vision field for ANPR solutions, especially, the solutions considering the recent innovations and enhancements in deep neural networks. Even though vehicle monitoring is a global problem, We will specifically focus on vehicle registration numbers issued by Azerbaijan and challenges regarding this selected domain. Given the domain, this paper will finally draw comprehensive comparisons between current ANPR technologies and derive meaningful conclusions based on data, experiments, and results