Fusing ladar and color image for detection grass off-road scenario

Abstract

It is necessary to extend the intelligent vehicle to navigate from structured environment to rough terrain, which is a great challenge for environment modeling. Ladar and camera are the most widely used sensors, but each of them has shortcoming. In this paper, SVM method is used to fuse the information from ladar and color camera. After registration, ladar point is represented by its position and neighbored pixels in the image. The height of the object as well as the H and S components of the color of the pixels are selected to represent the terrain. Grass and non-grass terrain are recognized based on the features. Experiment shows this method is simple and efficiency.