Automatic Rough Georeferencing of Multiview Oblique and Vertical Aerial Image Datasets of Urban Scenes. (September 2016)

• Record Type: Journal Article
• Authors: Verykokou, Styliani; Ioannidis, Charalabos
• Is Part Of: Photogrammetric record. Volume 31:Number 155(2016:Sep.)
• Page Start: 281
• Page End: 303
• Publication Date: 2016-09
• Subjects: computer vision -- georeferencing -- image matching -- oblique aerial images -- 2D transformations; Photogrammetry -- Periodicals; 526.982
• DOI: 10.1111/phor.12156
• Languages: English
• ISSNs: 0031-868X

Abstract

Abstract: Multi‐perspective airborne images that combine oblique and vertical views of the ground have proved to be a valuable source of information for numerous applications requiring a digital representation of the world. In this paper, an automatic methodology for rough georeferencing of large datasets of multiview oblique and vertical aerial images of urban regions without any metadata is proposed. Using feature‐based matching combined with robust model fitting and least‐squares techniques, the method requires the measurement of a minimum number of points with known coordinates in only one image. The results of this methodology are discussed through the presentation of a developed software suite which identifies the overlapping images, georeferences them, extracts their footprints, subdivides the images into groups based on these footprints and detects the images that cover a specific region.