Smart Motorcycle Accident Prevention System using Convolutional Neural Network

Abstract

The identification ofviolators on traffic rules isa highly essential but challenging task. A
surveillance system is mandatory to insist onroad safety measures, especially in two-wheeler motorists.
The system is capable of detecting non-helmet wearing persons and preventing deaths.This work
presented an automatic method for motorcycles detection on public roads, image handling technique, and
more advanced computer vision thatenhances machine learning classifier Convolutional Neural Network
(CNN) used to solve the problem. In the proposed system, the backgroundsubtraction technique used on
the surveillancevideo imagesto separate the subject from the surrounding object.Then the network is used
toclassify motorcyclists with and without a helmet.Then Hough transform edge detection techniqueis
applied to detect the number plate of the vehicle. Thus the information of the person without a helmet is
given to the control room through GPS. The personidentified without a helmet warnedandpunished with
penalty amount via the GSM system.The proposed system give an averageaccuracy rates of the
motorcyclist classified from other images is 93.94% and persons without helmet identification and
number plate detection were 93.25%