2015 GSA Annual Meeting in Baltimore, Maryland, USA (1-4 November 2015)

Paper No. 29-2
Presentation Time: 9:00 AM-5:30 PM

AUGMENTING GROUND-BASED 3-D MODELING WITH UAV PHOTOGRAPHY


CLARONI, Daniel J.1, CONNORS, Christopher D.1, HARBOR, David J.1, PFAFF, David2, BARNEY, Zachary1 and BALL, Stephen M.1, (1)Department of Geology, Washington and Lee University, Lexington, VA 24450, (2)Integrative and Quantitative Center, Washington and Lee University, Lexington, VA 24450, claronid17@mail.wlu.edu

Advances in Structure-from-Motion (SfM) computer vision techniques have revolutionized the ability to generate precise 3-D models of real-world objects, such as geologic outcrops. We present three case studies in the use of SfM to model at the cm-scale of exposures from Rockbridge County, VA. We used a high-resolution digital SLR for ground-based photo acquisition augmented with images acquired with an unmanned aerial vehicle (UAV), georeferenced with a combination of total station and differential GPS. The motivation for using both methods of photo acquisition was that the terrain was too rugged, or the objects of interest too complicated, for the photographer to maneuver to all orientations necessary for the SfM algorithms to sufficiently invert for 3-D positions accurately. Each of these three areas presented its own challenges. The first area was an abandoned quarry face that displayed a bedding plane that dipped 35 degrees and had almost perfect exposure over 700 m2. Features displayed include interference ripples and clear faults with both strike-slip and dip-slip displacement. The main struggles with conventional photography in this area were the pure size and slope of the face along with the concern of slope stability. Our second site was a bedrock riverbed. Here we were limited by the presence of the river and could not capture necessary perspectives of the site without the use of a UAV. Our third model location was a historical cemetery that has tall vegetation and monuments that we could only accurately capture if photos were taken from above. Preliminary analysis suggests precision at the sub-decimeter scale. For example, individual fault scarps in the quarry face that can be measured in the field having 1-2 cm offset can clearly be detected and measured in the models with good agreement. Our preliminary results suggest that although the semipro UAV is a very good solution to cover unreachable areas, there is no substitute for close up, high-resolution photos with a digital SLR. When possible, ground photos should be taken, but having a UAV to augment photo acquisition increased the overall coverage and quality of our models.