AI-based license plate detection of on-road vehicle

Acquiring images and detecting number plates of moving vehicles using motion platforms presents a challenge. To address this challenge, mobile cameras are utilized to capture license plate images of moving vehicles, which are then labeled and stored in a database. The test results obtained from our work, show a relatively high accuracy in terms of number plate recognition. The developed system achieves an accuracy of approximately 86% to 95% on our own datasets, depending on various weather conditions. The primary goal of our ANPR system is to achieve high accuracy in recognizing Odisha license plates. Other ANPR systems available use different types of pre-trained model methods, such as Alex Net and GoogleNet. Character recognition is also accomplished using various methods like Pytesseract. However, in this paper, we attempt to implement SSD Mobile Net (Single Shot Detector) and EasyOCR together. Our main objective is to demonstrate the use of a mobile phone instead of an expensive camera in number plate recognition, and therefore, we are interested in comparing the accuracy level achieved using a mobile camera. The results shows that using our scheme a good accuracy achieved using a smartphone camera.