THE USE OF STRUCTURE FROM MOTION TO GENERATE HIGH RESOLUTION TERRAIN MODELS

New photogrammetry techniques termed Structure from Motion (SfM) have emerged as a powerful tool for the analysis of geological features. This study presents 3D models made of bluffs from Wisconsin’s Lake Michigan coastline and glacial grooves from Kelleys Island, OH. These models were created from the combined use of drone photography, ground based photography and Agisoft Photoscan, a software that performs photogrammetric processing of digital images and generates 3D spatial point clouds. These technological innovations of SfM have created a new field of morphometric analyses in the geosciences.

Three-dimensional terrain models were created using SfM. For the glacial groove models a hand-held camera was used to photograph the area (15 x 100 m). The small area and subdued topography allowed for sufficient angles to be shot without the use of aerial photography. To create the bluff models photographs and video were shot from a drone at various altitudes along ca. 1000 meters of the bluff face. The optimum sets of photographs were produced by shooting 4k film and isolating still frames at intervals of 1.5 seconds – this method increased image overlap and subsequently the precision of the photogrammetric analysis. Images were processed with Photoscan in order to create 3D representations of the bluff faces and glacial grooves. Accurate (O(cm)), full color 3D point cloud representations of the bluffs and glacial grooves were generated with a standard desktop computer. In turn, digital elevation models and other interpretations of the data (such as seep locations and vegetation cover on the bluffs) were interpreted from the point clouds. The results of the DEM and other interpreted features from the bluff study were used as the basis for a hillslope stability model.

The minimal effort needed to generate 3D models using SfM renders it an appealing tool for the analysis of geological features. Combined with the use of drones SfM drastically reduces the time and cost limitations for generating spatial data, while simultaneously increasing precision and data versatility over LIDAR and other techniques.