Computerized Number Plate Detection and Generalized Character Recognition Using YOLOv8 Model

In recent years, Automatic Number Plate Recognition (ANPR) have been proposed and used in the car parking slots, highway monitoring system, traffic regulation etc. Detecting the number plate in the high traffic and also in dusty climate is a challenging work. Country like India with more population, it is not an easy task to identify the number plate with the recognition of characters irrespective of different size plate in the captured image. In this research, implementation of a State-of-the-Art YOLOv8 model for number plate detection and EasyOCR for character recognition. Dataset with number plate images is manually labelled and preprocessed before fed into the deep learning model. For the text localization and individual character recognition, detected Region of interest (ROI) is applied to preprocessing phase which includes Binarization, Denoising the image. Individual characters is extracted and the output is given in text format. The accuracy of this deep learning model is analyzed by Mean Average Precision (mAP) which is 93.7% and recall score is 94.7%.