New single camera vehicle detection based on gabor features for real time operation

Abstract

This paper presents a new novel approach for automatic vehicle detection from a live video based on texture Gabor features. Vehicle detection is a pivotal part in collision avoidance systems, blind-spot monitoring, and self-guided vehicles. This system uses a low cost camera mounted near the rear-view mirror to obtain the live video. The Gabor filter features have been used to identify the potential vehicles in the frames by using the support vector machine. The initial detection of potential vehicle candidates has been improved by using correlation techniques. A robust detection technique is developed in which the vehicles are detected with accuracy up to 9 % in day light image sequences. © 2011 Springer-Verlag.