Recognizing road environment for vision system on autonomous vehicles

Abstract

A road environment recognition system is presented. First, lane lines are detected based on our proposed Gaussian comparison function. Then a road prospective model is built on the basis of a new Gaussian semi-sphere approach to detect vanishing point. Using this model, three road parameters are computed, i.e. vehicle's transverse bias, yawing angle and the road slope angle. Obstacles such as vehicles in front of the current lane are also extracted. Simulation experiments prove the effectiveness of the system.