ABSTRACT
Assessing large-scale infrastructure systems and determining their health state under in-service or extreme hazard events post great challenges due to the less access in system level, particularly with the increase of their complexity. In this study, we explore the deep learning approaches integrated with computer vision for conditional assessment of large-scale systems through extracting critical information from the remote sensing data (e.g., satellite images or other resources). The original data were preprocessed on pixel level. The deep learning (R-CNN) model were selected and designed to gain each critical component condition. A full connected neural network was then used to assess the health state of the whole systems. The results demonstrated that the deep learning model accurately detected the damage location and level, while a full connected neural network effectively assisted the rapid identification and conditional assessment, thereby leading to high potential for large-scale bridges, and oil/gas pipeline assessment at both spatial and temporary scales over conventional methods, and particularly the deep learning quantitatively provided valuable information for infrastructure conditional assessment.
