Streamlined bridge inspection system utilizing unmanned aerial vehicles (UAVs) and machine learning.

• UAV is used to collect images for bridge condition assessment and heath monitoring. • Human-in-the-loop machine learning is used to segment elements from a point-cloud. • Damage detection from images is automated. • Damage growth is measured using images taken at different times/angles. • Damage location and extent are recovered by projecting 2D images to 3D point-cloud. Recently, the rapid development of commercial unmanned aerial vehicles (UAVs) has made collecting images of bridge conditions trivial. Measuring the damage extent, growth, and location from the collected big image set, however, can be cumbersome. This paper proposes a streamlined bridge inspection system that offers advanced data analytics tools to automatically : (1) identify type, extent, growth, and 3D location of defects using computer vision techniques; (2) generate a 3D point-cloud model and segment structural elements using human-in-the-loop machine learning; and (3) establish a georeferenced element-wise as-built bridge information model to document and visualize damage information. This system allows bridge managers to better leverage UAV technologies in bridge inspection and conveniently monitor the health of a bridge through quantifying and visualizing the progression of damage for each structural element. The efficacy of the system is demonstrated using two bridges. [ABSTRACT FROM AUTHOR]