License Plate Recognition System using Yolov5 and CNN
Shreya Raj, Yash Gupta, Ruchika Malhotra
Abstract
In recent years, many new applications of ANPR technology have been found. It is used in car parking management, towing systems, vehicle gate entry management, etc. Despite the usefulness of vehicle license plate detection technology, existing ANPR systems face many challenges to accurately identify number plates due to differences in size, orientation, the language of the license number plate across different regions throughout the world. In this research, we have implemented an ANPR system on a dataset that consists of number plates containing English alphabets and digits only. We have leveraged the state-of-the-art YOLOv5 object detection model for number plate detection. The output from detection is then processed and segmented into individual characters using image processing and contouring techniques. Then these individual characters are fed to a CNN model which returns the corresponding label. Finally, these character labels are merged to give the final output of the vehicle license plate number.