Vehicle detection in color images

Abstract

The AutoColor vehicle detection is a system that detects vehicles in images under different weather conditions. It first uses a novel color model of the road to classify each pixel lying on the road surface as either road or nonroad. The nonroad pixels are filtered and grouped into regions in the image. The regions are then back-projected to the ground plane and merged into vehicle rectangles. The AutoColor system is capable of detecting more than 90% of the vehicles under different weather conditions, including rainy and with strong shadows.