Night Time Vehicle Detection and Classification Using Support Vector Machine

Abstract

The paper presents a vehicle detection system by locating their headlights and tail lights in the nighttime road environment. The system detects the vehicles light in front of a micro CCD camera assisted vehicle i.e. oncoming & preceding vehicles. Our system automatically controls vehicle’s head lights status between low and high beams which avoids the glares for the drivers. The captured frames consist of number of bright objects over dark background. These objects are due to vehicle lamps, road reflection etc. The captured object features are used to train and classify the two classes of lights in vehicles light & other light source. The machine learning based approach, Support Vector Machine (SVM) is used to accomplish this task. The output of the SVM is simply the signed distance of the test instance from the separating hyperplane. The result show the SVM is effective to classify number of lights and it is useful for vehicle validation.