3D mapping for unmanned aerial vehicle combining LiDAR and depth camera in indoor environments

Abstract

Indoor reconnaissance missions for unmanned aerial vehicles (UAVs) pose significant challenges in scene reconstruction, mapping, and environmental feature extraction. Relying on a single type of sensor often results in limited accuracy, increased susceptibility to environmental noise, and a lack of comprehensive spatial information. To address these issues, this study proposes a mapping method that combines light detection and ranging (LiDAR) and depth camera data. The method collects data from both LiDAR and a depth camera integrated on the UAV, then performs preprocessing on both data sources to construct local 3D maps using the real-time appearance-based mapping (RTAB-Map) algorithm. Subsequently, the local maps are merged using a filtering method to generate a detailed and complete global map. Real-time experiments conducted on Ubuntu 20.04 using the robot operating system (ROS) Noetic libraries demonstrate that this multi-sensor fusion approach provides richer and more comprehensive environmental information, thereby enhancing the effectiveness of mapping tasks in unknown indoor environments.