Map-subtraction based moving-object tracking with motorcycle-mounted scanning LiDAR

This paper presents a method for tracking (estimation of position, velocity and size) of moving objects, such as cars, motorcycles, and pedestrians, in global navigation satellite systems (GNSS)-denied environments using light detection and ranging (LiDAR) mounted on motorcycle. 3D-point cloud environmental map is assumed to be acquired in advance. Distortion in scan data from the scanning LiDAR is corrected by estimating the pose (3D positions and attitude angles) of the motorcycle in a period shorter than the LiDAR scan period using normal distribution transforms (NDT) based map matching and the information from inertial measurement unit (IMU) via the extended Kalman filter (EKF). The corrected LiDAR scan data are compared with environmental map, and the LiDAR scan data of interest are extracted from the current LiDAR scan data based on map-subtraction method. The extracted scan data are mapped onto an elevation map, and moving objects are detected based on an occupancy grid method. Finally, detected moving objects are tracked based on the Bayesian Filter. Experimental results obtained in public road and university-campus road environments validate the effectiveness of the proposed method.