Vision-Based Analysis of Utility Poles Using Drones and Digital Twin Modeling in the Context of Power Distribution Infrastructure Systems

摘要

Wind-induced damages on leaning poles cause a large amount of losses and disruption to community functioning and interdependent infrastructure systems such as power networks. In order to reduce and prevent failure of utility poles in extreme weather events, traditional inspection techniques have been used to understand a condition of utility poles. Because of (1) uncertain detection rate depending on inspectors' visual observation skills and (2) labor-intensive process to inspect large-scale poles, camera-equipped drones have been leveraged to capture image data for reducing human errors as well as inspection time and increasing the number of inspections. Despite the benefits, prior works mainly focused on accurate detection and classification of utility poles, rather than pole inspection and maintenance through visual analytics. This paper presents a novel method to stream vision-based pole inspection information (i.e., maximum leaning angle of poles) into a virtual environment toward digital twin modeling in the context of power distribution infrastructure systems. By leveraging a single 2D image per pole obtained from a drone, the maximum leaning angle of utility poles over large areas are estimated. The outcomes of vision-based utility pole analysis are then fed into a virtual environment. For evaluation, case studies were conducted on Bryan, TX. The resulting digital twin model representing the stability information of utility poles has the potential to be used as a basis for better designing disaster preparedness and hazard mitigation strategies regarding the preventive maintenance of power distribution networks.