DAMAGE DETECTION AND LOCALIZATION IN MASONRY STRUCTURE USING FASTER REGION CONVOLUTIONAL NETWORKS

With current modern technology, manual on-site inspection can be assisted by automatic inspection, which is cost-effective, efficient and not subjective. In previous work, various image-based techniques have been applied to detect damages in heritage structures based on hand-designed feature extraction and classifiers. A heritage structure is composed of masonry walls, which are the components that are typically subjected to severe damages. This paper proposed a damage detection algorithm for a masonry structure based on Faster Region Convolutional Neural Networks (FRCNN). A labeled dataset for training the damage detection system in heritage masonry structure is created in this study, which is our first contribution as, currently, there is no public dataset available for masonry structures. The second contribution is the creation of a state of the art object detection system based on FRCNN for the detection and localization of damage in masonry structures. The results show that the proposed system performs well and can be used to detect damage in masonry structures with promising computational speed.