Chandler, Jim
3 CASE STUDIES
3.1 Tilting flume
The first projed was carried out on the 1.2m wide "Tilting Flume" locaed at Hydraulics Reseach, Wallingford, an
important center for fluvial research in the UK. The significance of the tilting flume projed was the ability to dredly
compare DEMs generated by digital photogrammetry and an independent laser profiling system. A small areaof the
existing flume bed (0.25 x 0.25m) was being measured using a Keyence (LC2450 laser displacement sensor, mounted
upon a horizontal motorized displacement frame, at a resolution of 0.5mm Each experiment lasted for three weeks,
with the bed surfacemeasured at the end of ead day, normally with a shallow depth of water («0.10 m) and negligible
flow rate. The disadvantage with the laser-based approach was the six hours required to measure the DEM combined
with its limited spatial coverage. Digital imagery of the bed was captured using the DCS460, the DCS460 and DCS420
synchronized and a Hasselblad semi-metric camera, in order to assssthe acuracy achievable using cameras of varying
type and inner geometry. The cameras were mounted on a frame 1.9m above the flume bed and 10 conventional control
points were placal on the flume edge and co-ordinated using theoddite intersection methods, The sensors were
cdibrated using bath in-situ and test-field self-caibration methods, bath methods being appropriate for establishing
camera inner orientation, (Chandler e al., in presg. DEMs were generated automatically at a resolution of 3mm over an
areaof 1.1 x 0.8m using OrthoMAX. Point elevations within the 0.25 x 0.25m area measured by the laser profiler were
. then downgraded to a resolution of 3mm and these 6889points were compared with the photogrammetric estimates to
obtain the r.m.s.e. accuracy assessments presented in Table 1. The results show that there is littl e variation with sensor
type, but that accurades were not high when compared with the
Sensor type r.m.s.e (mm) low camera/objed distance of 1.9m. Detail ed examination of the
DCS460 1.9mm surfaces derived showed that the major discrepancies occurred
DCS460DCS420 150mm between large particles in small regions of "dead ground" and so
Hassiblad TEE the size of the bed material in relation to the pixel scde is a
Ee dominant control (Sedion 4.0). However, the acaracy achieved is
Table 1. Accurades of sensor type acceptable for fluvial reseach and df greder significance was the
increased ared coverage (1.1 x 0.8m) and marked reduction in time required to oltain the imagery (only 10 minutes),
which all ows experimental work to conti nue.
Additional elements of research involved developing and assessing the potential of two-media photogrammetry and
deriving estimates of bed roughness from the digital elevation models. The two-media photogrammetry has allowed
imagery to be obtained through a shallow depth of water, without draini ng the flume and is reported upon more fully in
Butler et al., in press. Bed roughness has been quantified using a diverse range of methods, including texture extraction
using the Hurst operator, Fast Fourier Transforms and generation of semi-vari ograms and semi-variance surfaces. Semi-
variance surfaces have proved most valuable, because these have demonstrated that roughness can be parameterized at
different scales and can therefore be used to define boundary conditions for numerica flow models (Butler at al., 1998;
Butler & al, in press.
3.2 TheFlood Channel Facility
A second projed carried out at Hydraulics Research Walli ngford required generating DEMs of bed forms creaed in the
Flood Channel Fadlity (FCF). This comprised a large meandering sinusoidal channel with a wavelength of 15m and
channel width of 1.6m. The meandering pattern covered an area 15 x 8m maki ng efficient stereo-coverage difficult to
(mm)
Photogrammetry
-300 + a Point profiler (with sam pling points)
350 +
Figure 1, Accuracy of digital photogrammetry applied to the Flood Channel Fadi ty (Camera Height = 4.2m)
252 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000.