Vehicle Number Plate Detection and Recognition using YOLO- V3 and OCR Method

Abstract

Automatic Number Plate Recognition is one of the many information systems which is used data extraction from given vehicle image and apply the data for further usage in safe, secure and modernistic Transportation System. The Novelty of the project is that even if the image is blurred, our system can deblur the given image and apply it to the Machine Learning models further. In the proposed work, You Only Look Once [YOLO] V3 model for Region of Interest [ROI]; Convolution Neural Network [CNN] for optical character recognition was implemented. After the ROI is detected, it will be enhanced with pre-processing steps before it is fed to CNN model. Dataset of different Indian Number Plates’ Font was created, consisting of 6439 images of different alpha- numerical characters. Accuracy of 91.5 percent is obtained. The extracted and sorted characters of the number plate is cross- checked with the Indian RTO database and the information regarding which RTO the input vehicle image belongs to, is provided.