A SLAM METHOD FOR HANDHELD HEMISPHERICAL VIEW LASER SCANNING SYSTEM
Keywords: SLAM, 3D point cloud, iEKF, factor graph optimization, laser scanning system
Abstract. In view of the problem that some multi-line light detection and ranging (LiDAR) scans a small field of view in the vertical direction, a framework that uses an integrated handheld hemispherical view LiDAR and inertial measurement unit (IMU) scanning system for simultaneous localization and mapping (SLAM) is proposed. For the structural characteristics of the hemispherical view LiDAR scan lines, a ground segmentation pre-processing module based on seed points is designed. The segmented ground points are downsampled to eliminate redundant vertical constraints. The IMU data and the pre-processed point cloud are performed state estimation via a tightly coupled iterative Extended Kalman Filter (iEKF) to obtain the real-time poses. The detected loop closures provide global constraints for the point cloud map. The factor graph is used to process the back-end optimization incrementally to eliminate the accumulation error of the system. Data from diverse scenes are collected via a prototype system. Both qualitative and quantitative experiments are performed to prove the accuracy and performance of the framework. Experiments show that our framework outperforms the state-of-the-art SLAM methods for the hemispherical view LiDAR-IMU integrated scanning system.