Vision Based Damage Assessment of Stone Masonry Bridges Using Convolutional Neural Network

Stone masonry bridges, characterized by their age-old heritage, remarkable axial load-bearing strength, and exceptional design and construction, are a vital part of a nation’s historical legacy. Their preservation and functionality are of paramount importance, particularly in remote areas with limited resources. To achieve this, a thorough assessment of these bridges is essential, with a focus on early detection and classification of damages. This approach not only safeguards the knowledge embedded in these structures but also proves cost-effective when compared to building new bridges. In this study, an extensive analysis was conducted on 50 different masonry bridges, involving the collection of over 1000 images capturing both normal conditions and various types of damages. The major challenge in this endeavor lies in the accessibility of these bridges, as many are nestled amidst vegetation or surrounded by water bodies. To address this challenge, the research paper introduces a comprehensive investigation into the detection of damage in stone masonry bridges. It proposes the implementation of Convolutional Neural Networks (CNNs) to identify and characterize damage within these historic structures. The desired accuracy of the proposed study is targeted to 85%. Moreover, the scope can encompass the incorporation of Unmanned Aerial Vehicles (UAVs) to enhance assessment capabilities. This allows for the detection of various forms of damage and the overall condition of bridges. This research effort makes a substantial contribution to preserving and ensuring the ongoing use of stone masonry arch bridges, safeguarding their safety, functionality, and enduring cultural importance.