Coupling Deep Learning and UAV for Infrastructure Condition Assessment Automation

Abstract

We propose coupling the state-of-the-art computer technology deep learning and unmanned aerial vehicles (UAV) to automatically detect and assess the health condition of civil infrastructure such as bridges and pavements. UAV carrying high resolution camera and infrared thermography camera to collect a large amount of image data from the target infrastructure, which serves as inputs of trained deep neural networks for damage classification and condition assessment. Details of the framework that may guide the automation process are explained. We demonstrated preliminary application of using UAV and deep neural network in concrete crack and asphalt pavement distress classification. Challenges and needs for deployment of UAV and deep learning are briefly discussed in the end.