A UAV-based panoramic oblique photogrammetry (POP) approach using spherical projection

Abstract

Oblique photogrammetry is an emerging technology in unmanned aerial vehicle (UAV)-based low-altitude remote sensing, and oblique digital cameras are widely used for data acquisition. Moreover, panoramic cameras are becoming popular with their capacities of comprehensive range of imaging compared with original oblique cameras, a spherical panorama can be transformed to several perspective images which are similar to normal oblique images. In this paper, a UAV-based panoramic oblique photogrammetry (POP) approach is proposed to achieve georeferenced panoramic images and real 3D models together by using panorama image projection algorithms. First, a UAV-based panoramic imaging system with six fish-eye cameras is presented to obtain multiangle and large-field-of-view images to generate two-dimensional (2D) spherical panorama images. Second, two panorama image projection models from a 2D panorama to perspectives are proposed to regenerate small distortion images toward different directions as traditional oblique photogrammetry images. Third, aerial triangulation, dense matching and three-dimensional (3D) modeling are presented with these small distortion images to achieve panoramic oblique photogrammetry processing. Two real scenes in the city of Wuhan are taken to verify the feasibility and precision of the proposed approach. Experimental results show that the panoramic oblique photogrammetry for 3D modeling can achieve cm-level accuracy. The proposed approach can obtain aerial panoramas and georeferenced 3D models together, which can be an effective method for virtual geography environment visualization.