Sensor Pose Estimation and 3D Mapping for Crane Operations Using Sensors Attached to the Crane Boom

Abstract

This paper describes a method for sensor pose estimation, as well as creating large-scale 3D maps, for construction cranes equipped with a sensor system consisting of a camera, 2D lidar, and IMU. To tackle the challenges posed by the crane boom’s complex motion, we utilize an Extended Kalman filter (EKF) to improve the accuracy and reliability of sensor pose estimation. By combining pose estimates from Visual-Inertial Navigation System (VINS) with data from an additional IMU, we estimate the scale value of a monocular camera. This scale value, obtained from the EKF, is then integrated into the VINS algorithm to refine the previously estimated scale value. Slowly rotating 2D lidar is used to build a 3D map. Since there is limited overlap between 2D lidar scans, we leverage the estimated pose to align and construct a comprehensive 3D map. Additionally, we thoroughly evaluate the effectiveness of the latest VINS techniques, as well as the EKF-enhanced VINS approach, in the specific context of crane operations. Through comprehensive performance assessments conducted in both simulated and real environments, we compare the EKF-added VINS method with state-of-the-art VINS techniques. The evaluation results demonstrate that the EKF-added VINS method accurately estimates sensor poses, leading to the generation of high-quality, large-scale 3D point cloud maps for construction cranes.

Keywords

• 3D mapping
• 2D lidar
• IMU
• pose graph optimization
• complementary filter
• large crane