Assessment of geometric parameters of segmented crack on concrete building facade using deep learning

Abstract

Crack damage evaluation is a crucial index in building repair and maintenance for assessing the health of buildings. This study propose a framework for crack segmentation and damage evaluation to provide a rapid and accurate crack evaluation in practical applications. The framework consists of four steps: data set collection, training of the crack segmentation model, indicator establishment, and damage severity evaluation. Deeplabv3 + is used for crack segmentation. And different indicators are established based on connected domain analysis. To determine the severity of the crack, we establish a damage factor through hierarchical analysis. Field experiments are carried out on real buildings to assess crack damage and verify the method's feasibility and effectiveness. Our proposed method provides a fast and reliable way to evaluate cracks in buildings, which has a high potential for practical application.