A Global Photogrammetry-Based Structure from Motion Framework: Application in Oblique Aerial Images
Styliani VERYKOKOU, Charalabos IOANNIDIS
School of Rural & Surveying Engineering, National Technical University of Athens
In recent years, oblique aerial images have come back to the foreground, being involved in
various photogrammetric applications. Moreover, multi-camera systems have become a wellestablished
technology, providing oblique and vertical aerial images that depict both horizontal
and vertical structures of the environment from several perspectives, leading to an increasing
market availability of such kind of images. The main prerequisite for their metric exploitation
is the knowledge of the camera interior and exterior orientation parameters, which are usually
determined through a Structure from Motion (SfM) process that estimates corresponding
features between overlapping images and solves the multi-image aerial triangulation problem.
This paper presents a complete photogrammetry-based framework that solves the SfM problem,
which covers the topics of determining overlapping images, feature extraction, image matching,
rejection of outliers, feature tracking and bundle block adjustment. The proposed framework
adopts a global SfM workflow that relies on approximate camera exterior orientation
parameters, which are almost always available through the data provided by onboard GPS/INS
sensors. The proposed SfM methodology is applied in different configurations of oblique aerial
images (same perspective oblique images, multi-view oblique images and combined nadir and
oblique multi-view images) under a non-ideal aerial triangulation scenario characterized by lack
of well-distributed ground control points as well as minimum manual image measurements and
the results are outlined in the paper. The exterior orientation parameters computed through the
proposed SfM algorithm have better accuracy than the ones achieved through a commercial
SfM software package. Thus, the proposed global SfM framework proves to be a good
alternative solution to existing SfM methods.
Event: FIG WW 2019 Hanoi
Only personal, non-commercial use of this document is allowed.