Exploring Vehicle Number Plate Detection Methods: Yolo Models and Neural Network Models with Different Datasets

Abstract

Automatic vehicle number plate recognition (ANPR) is a significant traffic management solution that uses computer vision and deep learning techniques. With the rapid rise in vehicular traffic volumes worldwide, effective traffic monitoring systems have become imperative from safety and administrative perspectives. ANPR systems aim to automatically identify vehicle registration numbers regardless of changing light conditions. At the toll plazas and city entrances, ANPR assists with electronic toll collection and tracking vehicles moving in and out. It can also help with efficient parking management in congested urban areas. The core technical aspects involved are number plate detection (NPD) from images/videos captured by cameras, character segmentation to isolate the plate region, and optical character recognition (OCR) to read the alphanumeric characters accurately. State-of-the-art deep learning models like YOLOv7, CNN, LSTM networks and YOLOv5 are evaluated for their ability to detect plates under different circumstances. An infrared sensor facilitates clear number plate captures under varying light intensities. The overarching goal is to develop a robust ANPR framework through computer vision and deep learning for diverse traffic analytics and regulatory needs. The experiment result analysis of different deep learning models helps identify the best solution.