IEEE Geoscience and Remote Sensing Magazine - June 2018 - 13

rendering or visualization modes were tested. The digital
GeoloGiCAl interPretAtion
outcrops have been acquired through structure from moThe outcrop model used in this study is an important geologtion (SfM) using a global fusion method [14] that was then
ical-paleontological site located in Fazenda Arrecife in the
textured by a large-scale multiview stereo (MVS) technique
eastern part of Chapada Diamantina, State of Bahia, Brazil
[18] and finally validated with a georeferenced TLS point
(Figure 3). It occupies an area of approximately 7.5 km2, and
cloud (see Figure 5). In the field, it took approximately 1 h
its outcrops are used as analogs for oil exploration in terms of
to capture 324 photos, after which the reconstruction propotential reservoirs for hydrocarbon accumulation.
cess required 15 h on an Intel i7-3930K, Quadro 5000, and
This model contains exemplary exposures of bioherms
Tesla C2075. Moreover, another 10 h were spent manually
composed of columnar stromatolites from the Neoproterorefining the DOM.
zoic Era (the Salitre Formation) (Figure 4). These bioherms
A visual comparison was made to evaluate which modare associated with carbonate sediments of tempestites or
els offer better visual quality for geologists. The SfM model
storm deposits. Through MOSIS, we can observe the entire
along with a 3-D point cloud and mesh reconstruction
outcrop and its main structures, as well as the interactions
algorithms provided better preliminary results [13], [14],
between stromatolites and carbonate sediments. The stro[18]; however, to validate this approach, the scanned outmatolites range from 3 to 5 m in diameter and are usually
crop by TLS was used to verify the positional quality of the
subspheric to dome shaped. The measurement tool can be
former model.
used to precisely define their width and length. Another relBoth point clouds are compared through their posievant tool for outcrop model interpretation is plane drawtional discrepancies and standard deviation. A preliminary
ing and parametrization. One can select several points in
analysis has shown that the multi-images-based approach
the area and obtain the direction and dip for the best-fitting
can produce dense and visually accurate digital models (applane. It is also possible to create drawings on laminations,
proximately 0.5 million points). Comparing both point
contacts, or other structures of interest to the user. All data
clouds, a mean distance of 0.035531 m and a standard deviobtained by the user can be exported.
ation of 0.051424 m were observed. The meshes generated
The region where the outcrop is located is frequently
from multiview SfM, combined with the 3-D point cloud,
used for agriculture and cattle rearing, which will adversely
mesh reconstruction algorithms, and high-resolution texaffect the site in the future. In addition to its importance for
ture, provided better results in terms of visual quality, simteaching and use pre or post fieldwork, the DOM preserves
plicity, speed, and reasonable cost for fieldwork.
this impressive geological-paleontological site by digitally
preserving its smallest details.
Although a DOM of a relatively
small area was employed here, MOSIS
is not limited to this type of outcrop.
It also supports models acquired from
larger areas and from vertical outcrops.
These models also pose particular
challenges: as often happens in such
large-scale outcrops, an overview is
desirable. To achieve this, MOSIS al(a)
(b)
lows the user to inspect the models on
a table as though they were a replica of
the original. And, while different sizes
require different visualization techniques, the tools remain functional.
Some geological environments require
acquisition through drones or with
(c)
the aid of artificial lights. Such changes affect only the acquisition pipeline,
and the results may vary in quality and
data density, thus modifying the user's
interpretation capabilities.
evAluAtion AnD reSultS
To achieve more advanced visual results that allow geologists to approximate an experience similar to being
in the field, different data sets and
june 2018

(d)
FiGure 5. The DOMs with TLS. The image in (a) was obtained using photos and SfM/MVS

techniques; (b)-(d) show two clipped areas from TLS and SfM/MVS point clouds, respectively.
The point cloud (c) shows two holes (highlighted by red ellipses), while (d) does not.

ieee Geoscience and remote sensing magazine

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - June 2018