 |
| Titel |
From the air to digital landscapes: generating reach-scale topographic models from aerial photography in gravel-bed rivers |
| VerfasserIn |
Damià Vericat, Efrén Narciso, Maria Béjar, Alvaro Tena, James Brasington, Chris Gibbins, Ramon J. Batalla |
| Konferenz |
EGU General Assembly 2014
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 16 (2014) |
| Datensatznummer |
250094073
|
| Publikation (Nr.) |
 EGU/EGU2014-9411.pdf |
|
|
|
|
|
| Zusammenfassung |
| Digital Terrain Models are fundamental to characterise landscapes, to support numerical
modelling and to monitor topographic changes. Recent advances in topography, remote
sensing and geomatics are providing new opportunities to obtain high density/quality and
rapid topographic data. In this paper we present an integrated methodology to rapidly obtain
reach scale topographic models of fluvial systems. This methodology has been tested and is
being applied to develop event-scale terrain models of a 11-km river reach in the highly
dynamic Upper Cinca (NE Iberian Peninsula). This research is conducted in the background
of the project MorphSed.
The methodology integrates (a) the acquisition of dense point clouds of the exposed
floodplain (aerial photography and digital photogrammetry); (b) the registration of all
observations to the same coordinate system (using RTK-GPS surveyed GCPs);
(c) the acquisition of bathymetric data (using aDcp measurements integrated with
RTK-GPS); (d) the intelligent decimation of survey observations (using the open source
TopCat toolkit) and, finally, (e) data fusion (elaborating Digital Elevation Models). In
this paper special emphasis is given to the acquisition and registration of point
clouds.
3D point clouds are obtained from aerial photography and by means of automated digital
photogrammetry. Aerial photographs are taken at 275 meters above the ground by means of a
SLR digital camera manually operated from an autogyro. Four flight paths are defined in
order to cover the 11 km long and 500 meters wide river reach. A total of 45 minutes are
required to fly along these paths. Camera has been previously calibrated with the objective to
ensure image resolution at around 5 cm. A total of 220 GCPs are deployed and RTK-GPS
surveyed before the flight is conducted. Two people and one full workday are necessary to
deploy and survey the full set of GCPs. Field data acquisition may be finalised in less than
2 days. Structure-from-Motion is subsequently applied in the lab using Agisoft
PhotoScan, photographs are aligned and a 3d point cloud is generated. GCPs are used to
geo-register all point clouds. This task may be time consuming since GCPs need
to be identified in at least two of the pictures. A first automatic identification of
GCPs positions is performed in the rest of the photos, although user supervision is
necessary. Preliminary results show as geo-registration errors between 0.08 and
and 0.10 meters can be obtained. The number of GCPs is being degraded and the
quality of the point cloud assessed based on check points (the extracted GCPs). A
critical analysis of GCPs density and scene locations is being performed (results in
preparation). The results show that automated digital photogrammetry may provide
new opportunities in the acquisition of topographic data at multiple temporal and
spatial scales, being competitive with other more expensive techniques that, in
turn, may require much more time to acquire field observations. SfM offers new
opportunities to develop event-scale terrain models of fluvial systems suitable for
hydraulic modelling and to study topographic change in highly dynamic environments. |
|
|
|
|
|
|