Plaka tanıma sistemleri ve hibrit bir sistem önerisi

Abstract

Today, traffic safety is gaining more and more importance due to the increase in the number of vehicles and traffic density. In this context, license plate recognition systems are vital for traffic control, parking control, security and many other applications. License plate recognition is the process of converting images of license plates into text. This process is particularly useful in applications such as vehicle recognition or tracking. License plate recognition systems are primarily used for recognizing or detecting vehicles. These systems use a set of algorithms to recognize and understand the text on the license plate, primarily by processing a license plate image taken by a camera or imaging device. In this study, it is focused on how license plate recognition systems work in different countries and languages, what algorithms and techniques are used and how these systems are created. In addition, the latest developments in license plate recognition systems and the direction in which future research can move are discussed. The aim of the study is to provide an overview of how license plate recognition systems work and to present an innovative method that includes the use of modern technologies. In the study, CNN, RNN, SSD algorithms and three YOLO versions were used for the detection of license plates. After the plate detection stage, the characters were read with the OCR model. The methodology of our proposed model was compared with the literature and successful results were obtained. The Precision value is 91%, the Recall value is 82%, at 0.5 the MAP is equal to 89%, and at 0.5:0.95 it is 88%, and the Execution time is 0.173 seconds.