HELMET DETECTION AND LICENSE PLATE RECOGNITION USING CNN

Abstract

Two-wheelers are the popular mode of transport in the country. Due to its increased usage, the rate of
two-wheeler accidents has also raised significantly. According to a recent survey, the majority of motorcyclists
do not wear helmets, which accounts for approximately 32% of two-wheeler accidents. This issue should be put
forth to the people, to understand the seriousness. As a result, the government has deemed riding a motorcycle
without a helmet a criminal violation. Manual systems were implemented to perform the identification of people
not wearing helmets,however, it was not efficient at all times since it was a tedious process. Thus, automation of
this system is required. The existing automated approach provides a low accuracy. In this paper, we propose a
convolutional neural network(CNN) based approach for automatically detecting helmets from real-time
surveillance videos. The frames from the surveillance video are initially acquired and the key frames are
extracted in the suggested approach. Then these key frames are classified as motorcycles or not, using CNN
method. Then the motorcyclists without the helmet are identified using the CNN method. Lastly, the violator’s
license plate’s characters are recognized using Support Vector Machine(SVM) classifiers and is stored in the
database for generating and sending the fine amount to the violators through SMS using the Twilio API.