Vehicle Front Environment Recognition Using Stereo Vision

Abstract

In this paper, we propose vehicle front environment recognition method using stereovision system. In the vehicle, which we developed, two cameras are installed in front of rearview mirror and a disparity image is obtained by analyzing grabbed stereo-image. The road plane is extracted by simple geometrical projection of disparity image with an assumption that the stereo camera is installed in parallel with road plane. Then, obstacles are recongnized by computing the ratio of the object that exists in the fixed height from a road surface. As a result of implementing on general-purpose PC and performing an examination on a highway, it was confirmed that real-time computation is possible for the proposed technique, and it has very high robustness.