Structural Damage Detection, Localization, and Quantification via UAV-Based 3D Imaging

Abstract

Visual damage inspection for civil structures is a labor-intensive and timeconsuming task. We propose an autonomous UAV-based pipeline for crack and spalling detection, localization, and quantification. Through fusing 3-dimensional (3D) reconstruction and 2D damage detection after performing UAV-based imaging for an engineering structure, the process generates a damage-annotated 3D information model with rich metadata, including the size and type of damage and its location relative to the structure. The pipeline is composed of four steps: image acquisition via UAV, 3D scene reconstruction, crack/spalling detection and extraction using a deep neural network, and 3D damage localization and quantification. To validate this process, UAV images from three full-scale concrete columns are processed, and results are evaluated in this paper. The results demonstrate that the proposed pipeline can provide accurate and informative 3D condition mapping for civil structures. The authors envision that by employing this UAV-based automatic process, structural damage inspection can be conducted much frequently and rapidly with a significantly low cost.