On-site Semantic Mapping and Waypoint Planning for Autonomous Aerial Bridge Monitoring

Abstract

Effective monitoring of aging bridges is critical for ensuring their safety and maintenance. This study introduces a framework for on-site autonomous aerial bridge monitoring using sensor fusion and SLAM (Simultaneous Localization and Mapping). The proposed method utilizes a lightweight LiDAR sensor and a mini PC onboard a drone to generate real-time 3D semantic maps and flight waypoints. YOLOv8-based image segmentation is employed to identify bridge components, achieving a mean Average Precision (mAP50-95) of 86.6% across test data. Segmentation requires less than 10 milliseconds per frame, while processing LiDAR point clouds takes less than 1 second per frame. Waypoint generation based on the semantic map is completed in under 3 seconds. These results demonstrate the framework's capability to deliver precise and reliable on-site monitoring. This system provides a significant advancement in autonomous aerial bridge inspection by enabling efficient and real-time operation.