Automatic Number Plate Recognition using YOLOv5
Keywords:Optical Character Recognition (OCR), Automatic Number Plate Recognition (ANPR), YOLO, Real-Time Object Detection System, Convolutional Neural Network (CNN)
Automatic number plate recognition (ANPR) is a system that uses a vehicle's number plate to identify it. The popularity of ANPR technology has increased over the past years as a result of its variety of applications in different fields. To build the ANPR system two stages are considered. In the first stage, to segregate the number plate from the rest of the image, we utilize the pre-existing open-source system YOLOv5. You Only Look Once (YOLO) is a state-of-the-art, real-time object detection algorithm that utilizes regression to predict bounding boxes and associated class probabilities for the whole image using a single convolutional neural network (CNN). In the second stage, once the number plate is detected, it is cropped from the image and is passed on to detect characters using optical character recognition (OCR). OCR identifies text inside an image file and converts it into a machine-readable text form to be used for data processing. For license plate recognition, compact classifier models combined with YOLOv5 are experimented from the perspective of obtaining high efficiency of implementation in Raspberry-Pi embedded systems. Based on investigated compact solutions analysis, we determine the optimal solution with maximum accuracy in terms of minimum response time. The solution is integrated into the Raspberry Pi platform together with software modules that allow access via the Internet and the management of the database containing the recognized numbers. This paper aims to create an optimal ANPR system for embedded systems by implementing a solution that increases the performance, efficiency, scalability and accuracy of automatic license plate recognition software.
Copyright (c) 2022 Agnia CODREANU, Razvan Alexandru BRATULESCU, Pavel MURESAN, Mari Anais SACHIAN, George SUCIU
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.