Page 6 - i1052-5173-26-8
P. 6
Figure 2. Pseudotachylite specimen from Vredefort impact structure. ©2016 Google Inc. Image: Landsat. Map by Hartwig Frimmel. Online version can be rotated,
and is available at http://dx.doi.org/10.1130/GSATG257.S2.
GSA TODAY | AUGUST 2016 2016; TurboSquid, 2016]) was used to clean scanning errors and into Google Earth or Cesium virtual globes unless they evolve in
reduce model size. Google Earth literally shreds models with tandem, but will be accessible from virtual field trip stops via
more than 64,000 vertices, so reducing the number of vertices is HTML hyperlinks to modern browsers (Gemmell, 2015), of which
essential for most raw scans. Of the many vertex reduction the fastest appears to be Waterfox (2016).
options in MeshLab, the only one that worked whilst main-
taining specimen quality was Quadric Edge Collapse Decimation VisualSFM (Wu, 2013) is an open-source application with
(see the GSA Supplemental Data Repository [footnote 1]). The enhanced SfM editing capabilities; however, it requires
model was exported from MeshLab in COLLADA format for use command-line competency and is not for the faint-of-heart.
with Google Earth. PhotoScan from Agisoft (2016) is a more popular choice (Pitts et
al., 2014; Shackleton, 2015) and whilst not free, is deeply
MULTI-VIEW STEREO AND STRUCTURE FROM MOTION discounted for education. Bemis et al. (2014) review other SfM
methodologies, including UAV outcrop mapping. Probably the
The most exciting recent modeling innovations are in the field easiest SfM application for beginners, however, is Autodesk’s
of multi-view stereo (MVS) photogrammetry. Se and Jasiobedzki 123D Catch.
(2008) used video imagery from an unmanned vehicle and the
Simultaneous Localization and Mapping (SLAM) algorithm to Schott (2012) modeled mud cracks using Autodesk’s original
monitor an active mine. An algorithm called Structure from SfM application, PhotoFly—since renamed 123D Catch—which is
Motion (SfM) uses multiple still images from a smartphone or freely available from Autodesk (2016; there is a premium version
other digital camera to build 3D models. Snavely et al. (2008) and with a US$10 monthly fee). Karabinos (2013) used it to create
Enqvist et al. (2011) developed non-sequential SfM, enabling outcrop and boulder models. De Paor (2013) described the process
model construction from image searches (Schonberger et al., of porting 123D Catch models to Google Earth by processing
2015). However, Sakai et al. (2011) require only two photographs, them through MeshLab. Bourke (2015) used SfM to model an
and Gilardi et al. (2014) created 3D beach pebbles from a single indigenous Australian rock shelter; Lucieer et al. (2013) mapped
orthogonal photograph. landslide displacement using SfM and UAV photography; and
MCG3D (2015) made particularly good use of annotation capa-
The bleeding edge of SfM technology is Autodesk® Memento bilities in a geo-tourism application.
(2016), which at the time of this writing was in public beta-test
phase. It was slated for commercial release in May 2016 under the Figure 3 shows a mantle xenolith from Salt Lake (Āliamanu)
new name Autodesk® ReMake. It promises to accommodate Crater, adjacent to Pearl Harbor, Hawaii. The verilith was
billions of vertices with no limit on the number or resolution of collected by Michael Bizimis, University of South Carolina, and
images. Such models will doubtless be too large to embed directly mailed to the author for SfM modeling. Because the most impor-
tant part of this specimen is the saw-cut surface, it was possible to
6