Currently, necessary bridge inspections are mostly performed manually by highly skilled employees who have years of experience in detecting potential dangerous damage, such as cracks on bridges. However, these manual inspections, although considered reliable, are still subjective and usually time-consuming as well as dangerous for the employees.
In this subproject of the SPP, we will combine several state-of-the-art sensor technologies and emerging approaches to data processing and machine learning: Unmanned aerial vehicles (UAVs) - or drones - will be equipped with high-quality cameras and used to access hard-to-reach areas of the bridge and capture high-resolution images that will be processed by a computer vision module for automated analysis, making them a powerful and flexible tool for structural health monitoring. We will use high-resolution full-frame metric cameras in combination with appropriate multirotor systems to monitor both comprehensive, whole-structure geometric deformations and visual damage to the structure. In addition, we will use structured light scanners that can resolve a surface three-dimensionally in the submillimeter range. All sensors will be registered in a coordinate system defined by the infrastructure and stable in time. By using an innovative approach to control the UAV and combining all sensor data in an advanced deep-learning image interpretation approach, we propose a holistic image-based monitoring method.
Within the priority program, this subproject is part of cluster E "Data-driven methods". The methods to be developed will be applicable not only to bridge structures but also to the monitoring of other complex infrastructure structures, as they are generic in principle.