UAV-based automated 3D condition detection of railroad crossties

Abstract The railroad industry plays a crucial role in freight and passenger transportation in the United States. However, the safety of this system is compromised by frequent train accidents, many of which are caused by damaged crossties. Traditional crosstie inspection methods including manual checks and sensor-based monitoring are labor-intensive, costly, and limited in scope. Currently, there is no remote sensing based method monitoring various crossties conditions without interrupting existing railroad operations. Unmanned aerial vehicle (UAV)-based photogrammetry offers an efficient, cost-effective and non-contact solution for large-scale inspections, enabling comprehensive 3D reconstruction of crossties made of different materials. This paper proposes a novel remote sensing approach for crosstie inspection via integrating UAV-based sensing with photogrammetry techniques. The framework consisting of data collection, 3D reconstruction, and post-processing to identify and analyze crosstie conditions is introduced. A series of laboratory and field experiments adopting this framework detecting different materials (e.g., concrete, steel and wood) crossties conditions have been conducted. This is the first time adopting UAV-based photogrammetry detecting full-scale railroad crossties conditions. The crossties detection results achieve 90% accuracy and 1.0 recall conditions from a large group of crossties in the field. These results demonstrate that our UAV-based photogrammetry method is promising for real-world railroad crossties inspection and maintenance.