Real-Time Intersection Detection Based on Satellite Image and 3D LIDAR Point Cloud

Real-time detection of road intersections is an essential task for unmanned vehicle, especially when driving in an unknown environment with GNSS signal blockage. This paper proposes a novel intersection detection method by fusing information from LiDAR point cloud and satellite image. Firstly, based on the traditional laser beam model, a multi-hypothesis intersection detection method is proposed. Then, we extract intersection templates from the satellite image. These intersection templates serve as the prior for the intersection detection algorithm and significantly reduce the false positive rate. The experimental results on KITTI-odometry dataset show that our method can automatically identify all intersections on the target path.