Deep Learning-Based Automated Background Removal for Structural Exterior Image Stitching

This paper presents a deep learning-based automated background removal technique for structural exterior image stitching. In order to establish an damage map of structure using unmanned aerial vehicle (UAV), close-up vision scanning is typically required. However, unwanted objects are often captured within the scanned digital images. Since unnecessary cause serious distortion on stitching process, they should be removed. this paper, depth estimation proposed. Based fact that region interest has closer working distance than ones from camera, images can automatically removed network. addition, optimal selection based feature matching-based overlap ratio The proposed experimentally validated UAV-scanned acquired in-situ high-rise building structure. validation test results show obtained produce precise with computational cost reduction 85.7%, while raw fail construct and distortion.