Detecting Approaching Objects at Intersection Using on-Vehicle 3D-LiDAR for Automated Driving Vehicle

In our research, we developed a method for detecting approaching objects at intersection by focusing on geometric features of point cloud obtained from 3D-LiDAR, without using pre-generated maps to understand the environment. This method can be applied to intersection with diagonal crossings, and can detect approaching vehicles and pedestrians at distances of up to 49 m and 38 m, respectively. The results also showed that the detection was robust and continuous. Furthermore, this process can be used in 50 ms per a frame, so that can be used in real time. This will lead to collision prediction and judgment of starting for automated vehicles.