Feature-based image stitching for panorama construction and visual inspection of structures

This study presents a feature-based image stitching method with multi-level constraint criterion for panorama construction and visual inspection of building structures. The comparison of global view and local resolution over building exterior is discussed regarding practical implementation. An inspection-oriented methodology framework with optimized inlier distribution is designed for generating a feasible and reliable building panorama by using ordinary optic images. Two illustrative examples, including an earthquake-damaged masonry wall and a high-rise building with stone curtain walls, are experimentally investigated. The severely developed structural crack is fully mapped with stitched image and extracted in preparation for further quality evaluation. The curtain wall of the high-rise building is successfully constructed by using UAV-based images. The panorama quality is further compared with commercial stitching software and several improvements are illustrated in the particular case. In addition, the reliability of the proposed feature-based stitching approach is parametrically studied with different setups of input images.