Night Time Headlight Detection using CNN Based Object Tracking

Abstract

Due to bad visibility many accidents take place at night time. High beams used by oncoming vehicles produce glare and pose discomfort to people, thereby contributing to a big portion of these accidents. Our main goal is to detect and track the oncoming vehicle's headlights from the images extracted from a camera by using a trained CNN model and switch the lighting of the vehicle from high beam to low beam. When there is no oncoming vehicle, the lighting automatically switches to high beam. This will reduce the discomfort caused to the oncoming vehicle's driver and improve visibility for both the vehicles greatly, thereby reducing the risk of an accident.