Mosaicking UAV orthoimages using bounded Voronoi diagrams and watersheds

Abstract

ABSTRACTPanoramic mosaicking is a key step in the process of aerial photogrammetry, which combines multiple images to form a single image with a wide field of view. In this article, a refined global seam-line network generation method based on bounded Voronoi diagrams and watershed segmentation algorithm is proposed for mosaicking unmanned aerial vehicle (UAV) orthoimages. First of all, valid range polygons and image centres are extracted from orthoimages. Second, an initial seam-line network is generated by a bounded Voronoi diagrams and polygon Boolean operations. Third, the seam-line network is refined by marker-controlled watershed segmentation algorithm. Finally, a photomosaic is produced by multiband blending mosaic. The panoramic mosaicking approach is comprehensively discussed in terms of data volume and degree of overlapping. Experimental studies performed on several groups of UAV orthoimages show that the proposed approach provides a feasible and promising solution to large-scale UAV photomosaic generation.