Lidar-camera Based 3D Obstacle Detection for UGVs

A 3D obstacle detection method based on lidar and camera fusion was studied for the safety and autonomy of unmanned ground vehicles. Traditional 3D object detection methods usually rely on a single sensor, such as directly segmenting an object on point cloud or image. We present a method combining lidar and camera by projecting the lidar points onto the camera and fusing the image information of the neighborhood, then a high-resolution depth map is formed based on the accelerated high-dimensional filtering method. By detecting obstacles on the depth map instead of directly acting on the point cloud, the speed of obstacle clustering is increased. And the detection accuracy is improved due to the improvement of the resolution. We validate our results on the kitti dataset, showing the adaptability of our approach.