*
(a) The UAV-MVS point cloud below Poisson (blue) and TIN
(light blue) strips.
(b) The T-MVS point cloud below Poisson (brown) and TIN
(pink) strips.
(c) The T-MVS point cloud below UAV-MVS Poisson (blue) and
T-MVS Poisson (brown) strips.
Figure 9: 6 cm wide strips of Poisson and TIN surfaces viewed
over a pebbly beach section of the points clouds from which they
were derived (Z-10 cm), each natural coloured dot has a 14mm
diameter.
Two datasets were derived using the technique, one using terres-
trial photography and the other using photography acquired via an
unmanned aerial vehicle (UAV). The two point clouds provided
dense point coverage of the areas captured in the imagery, the ter-
restrial MVS dataset had 3-5 points per cm? and the UAV-MVS
dataset had —1-3 points per cm?. Once georeferenced the two
clouds coincided quite well, however in future studies compar-
ison will be between coregistered datasets. Triangulated mesh-
ing and Poisson surface reconstruction was used to create surface
models and these models were compared and evaluated to assess
how well the terrain and surface features were portrayed. The
point clouds produced using MVS have point normals associated
with each point and this allows detailed surface features to be de-
rived using Poisson surface reconstruction. The derivatives that
can be extracted from such a detailed surface representation will
benefit from the Poisson algorithm as it combines global and local
function fitting and seems to smooth the data and the process is
not strongly influenced by outliers in the point cloud. Future stud-
ies will undertake quantitative assessment of the differences and
evaluate the potential of these techniques for change detection,
in this study area the fine scale coastal erosion that is occurring
may be indicative of climate change and, if this technique proves
useful, UAVs may be a viable tool for focussed monitoring stud-
ies. The issues faced in vegetated areas and areas with complex
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
geometry that result in sparse patches in the point cloud need to
be investigated, it may be that the key areas of change are still
well represented. The MVS technique has a great deal of poten-
tial both in natural and man-made landscapes and there are many
potential applications for the use of UAVs for remote sensing
data capture, alongside laser scanning and more traditional tech-
niques, to provide very detailed and precise representations of a
range of landscapes at key moments. Application areas include
landform monitoring, mine surveying and other environmental
monitoring. Qualitatively, the outputs from the UAV-MVS pro-
cess compare very well to the terrestrial MVS results. The UAV
can map a greater area faster and from more viewing angles, it
is therefore an ideal platform for capturing very high detail 3D
snapshots of these environments.
ACKNOWLEDGEMENTS
The authors would like to thank Darren Turner for his logistical,
technical and programming support and UAV training. Thank
you to Myriax for the scholarship license for the Eonfusion nD
spatial analysis package. In addition, for making their algorithms
and software available, we would like to give thanks and appre-
ciation to Noah Snavely and his team for Bundler, David Lowe
for SIFT, the libsift team for SIFTFast, Yasutuka Furukawa and
Jean Ponce for their multi-view stereopsis algorithms (PMVS2
and CMVS) and Martin Isenburg for LASTools.
References
Bolitho, M., Kazhdan, M. and Burns, R., 2009. Parallel poisson
surface reconstruction. Advances in Visual Computing LNCS
5875, pp. 678—689.
Harwin, S. and Lucieer, A., 2012. Assessing the accuracy of
georeferenced point clouds produced via multi-view stereopsis
from Unmanned Aerial Vehicle (UAV) imagery. Currently in
review, publication pending.
Jazayeri, L, Fraser, C. and Cronk, S., 2010. Automated 3D Object
Reconstruction Via Multi-Image Close-Range Photogramme-
try. International Archives of Photogrammetry, Remote Sens-
ing and Spatial Information Sciences XXXVIII(2000), pp. 3-
8.
Kazhdan, M., Bolitho, M. and Hoppe, H., 2006. Poisson surface
reconstruction. In: Proceedings of the Eurographics Sympo-
sium on Geometry Processing.
Lim, S. and Haron, H., 2012. Surface reconstruction techniques:
a review. Artificial Intelligence Review.
Lowe, D. G., 2004. Distinctive Image Features from Scale-
Invariant Keypoints. International Journal of Computer Vision
60(2), pp. 91-110.
Mikrokopter, 2012. WiKT”
http://www.mikrokopter.com.
"MikroKopter
Remondino, F. and El-Hakim, S., 2006. Image-based 3D Mod-
elling: A Review. The Photogrammetric Record 21(115),
pp. 269-291.
Snavely, N., Seitz, S. M. and Szeliski, R., 2006. Photo Tourism:
Exploring image collections in 3D. ACM Transactions on
Graphics 25(3), pp. 835.
Zhou, K., Gong, M., Huang, X. and Guo, B., 2010. Data-Parallel
Octrees for Surface Reconstruction. IEEE transactions on vi-
sualization and computer graphics 17(5), pp. 669—681.
