Automatic number plate detection using TensorFlow in Indian scenario: An optical character recognition approach

Abstract

India, a country of over 130 Crores population, witnessed an exponential rise in the number of vehicles running on the road. As per the survey made in the year 2019, in India there is one death in every 4 min occurring due to over speeding. Moreover, there are around 2 lakhs vehicles being stolen in India per Year. This number can be reduced to a greater extent if proper steps are taken. Automatic number plate detection is a very important tool when it comes to detecting the license plate of a vehicle from the image of a vehicle so taken. Moreover, same tool can be implemented in the case of over speed monitoring, smart parking management system, law enforcement and electronic toll collection. In this paper, work has been carried out to automatically detect the number plate of a vehicle using computer vision technique. TensorFlow, a model developed by Google is used for the purpose of training and testing of the algorithm. A data set of about 500 images obtained at the different condition of the illumination, camera angle and varied distances has been used while training and testing the model. This resulted in an improved accuracy of about 85% with which the license plate has been detected. Further optical character recognition is implemented in order to get the license number from the image being detected.