Real-Time Detection of Vehicle License Plates Numbers

Abstract

Object Detection (OD) techniques have emerged as the key to dealing with the most complex computer vision problems in recent years. Vehicle License Plate Detection (VLPD) is the most important stage of any vehicle license plate recognition system (VLPR) because changes in its size, orientation, color, and background, contrast, and resolution have a direct impact on the system’s robustness and accuracy. The purpose of this paper is to present an object detector for detecting vehicle license plates in real-world scenes. We developed a new dataset of vehicle license plate numbers and used it to train our custom model. In YOLO-v3 layers, we decreased the number of classes to one in order to improve the detector. When we evaluated the system, we achieved precision, recall, and overall accuracy metrics of 0.95, 0.96, and 92.83 percent, respectively.