Indian Vehicle Number Plate Detection and Recognition using Deep Learning

Abstract

The increase in the number  of  vehicles  in  the  last few years has made it challenging to manually note the number plate text of the vehicle. Hence, in order to reduce the manual work, there is a need to propose a methodology that can detect the number plate region from the input image and recognize the characters of the number plate. Systems have been built for the same using Image Processing techniques, but this technique fails to provide accurate results occasionally in the case of real data. Modern technology such as Deep Learning overcomes this problem. Hence, a deep learning-based methodology is proposed to detect the number plate region from the input image and recognize its characters. Using the Region-based Convolutional Neural Networks (RCNN), the number plate region is detected and using Convolutional Neural Networks (CNN), the characters are recognized from the detected plate region. Once the characters of the number plate are obtained, the system derives the name of the state to which the vehicle belongs, and using the Vahan-info website, the complete details of the vehicle are obtained. The system also stores the number plate text with its state name into the database to maintain a record of number plates detected. The proposed system provides promising results, resulting in an accuracy of 98.46% for RCNN model and an accuracy of 95.98% for CNN model.