ul 2004
ible for
ination
mbina-
of the
of ap-
'ompo-
greater
sfy the
ce dif-
sidered
of each
ON
proved
" preci-
nagery
vithout
GCPs.
of the
dataset
d zero,
ulation
1viron-
ent for
s were
] each
nt pre-
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
Scene Block I (four strips) oo (um)
combination | 0 GCPs | 4 GCPs | 9 GCPs | 12 GCPs
F/N 33 3.4 3.3 3.4
F/B 4.1 4.1 4.3 4.2
N/B 2.8 2.8 2.9 2.9
F/N/B 3.4 3.4 3.4 3.4
Table 3. Phase II ORIMA 69 results
It is noted that importing GCPs in different patterns had only a
slight effect on the adjustment precision of the two scene block
combinations. Triangulating three panchromatic scenes pro-
duced a weighted average of the two scene adjustments.
4.4 Stereo measurement tests
The independent check points described in Figure 2 were meas-
ured in SOCET SET. Stereo measurements were made to 24
combinations of ADS40 images (the F/N/B triangulation has
three stereo possibilities, namely F/N, F/B and N/B). A sum-
mary of the results from the stereo measurements is given in
Appendix B. The importing of GCPs into the triangulation
process resulted in an improvement in the accuracy of meas-
urement. The check points in the F/N/B triangulation were
measured in the three stereo combinations to determine which
stereo possibility should be used to achieve the optimal result.
It is clear that triangulated combinations without GCPs returned
the lowest accuracy, particularly in the vertical component. The
importing of four GCPs into the bundle adjustment improved
the accuracy significantly in all combinations. The influence of
importing one, two and three GCPs to the triangulation process
is therefore now being investigated to see if this eliminates
what is believed to be a “datum” problem with the vertical
component. Increasing the GCP count to nine improved the
measurement accuracy further. In the case of the 12 GCP con-
figurations, no improvement was made to the measurement ac-
curacy, indeed a slight degradation was evident from the
graphs. This is suspected to be as a result of an erroneous GCP,
but further investigations are ongoing.
It can be summarised that for optimum results, all three pan-
chromatic scenes should be used in the triangulation process.
The F/B stereo combination produced the optimum stereo
measurements and the use of nine GCPs in the block provided
the best solution for this phase of testing.
5. PHASE III: BLOCK II, POST-TRIANGULATION
5.1 Block II triangulation
In Phase III, the two cross strips were added to the adjustment
so that six strips were used in the triangulation. The procedures
followed in phase II were repeated in order to study the impact
of importing cross strips into the triangulation together with dif-
ferent GCP patterns. The same panchromatic scene combina-
tions as described in Section 4.1 were adopted.
5.2 Ground Control point configurations
The GCPs configurations, selected as subsets from the provided
dataset, comprised zero, four, nine and 12 GCPs and were lo-
cated as shown Appendix A (Figures A5 to A8).
5.3 Results from Block II adjustments
Each ORIMA output was analysed after running CAP-A and
arcas of weakness identified to fulfil the required criteria. The
final results (o9) of CAP-A adjustments are listed in Table 4.
Scene Block II (six strips) 0 (um)
combina-
tion 0 GCPs | 4 GCPs | 9 GCPs 12 GCPs
F/N 3.8 3.6 3.6 3.6
F/B 4.5 4.3 4.3 4.3
N/B 3.3 32 32 3.2
F/N/B 3.8 3.8 3.8 3.8
Table 4. Phase III ORIMA op, results
It is noted that adding cross strips produced a slightly degraded
precision in comparison with Phase II (Table 3). Importing
GCPs in different patterns again had minimal effect on the ad-
justment precision of different scene combinations.
5.4 Stereo measurement tests
The stereo observation process used in phase II was repeated
for the Block II (six strips) adjustments. The same independent
check points were used in the stereo measurement test. Appen-
dix C summarises the results of the 24 sets of measurements. In
this case it is noted that adding cross strips without GCPs did
not significantly improve the vertical stereo measurements ac-
curacies. Therefore, importing GCPs is still required to maxi-
mise the height accuracy.
6. CONCLUDING REMARKS
The results presented in this paper show that, despite the pres-
ence of GPS/IMU data, the digital aerial triangulation of
ADS40 images still requires some GCPs in order to maximise
photogrammetric precision and accuracy. All three panchro-
matic scenes (F, N and B) should be used in triangulation to
achieve optimum results. For optimum stereo plotting results,
the F/B panchromatic scenes (comprising the largest stereo an-
gle of all scene combinations) should be adopted. Finally, the
addition of cross strips minimizes the number of GCPs required
to four in the corners of the block. Adopting these guidelines
resulted in a near-homogeneous stereo measurement accuracy
of 1 pixel in X, Y and Z axes (with measurements undertaken
by a relatively inexperienced photogrammetric operator in the
form of a research student).
REFERENCES
Sandau, R., Braunecker, B., Driescher, H., Eckardt, A., Hilbert,
S., Hutton, J., Kirchhofer, W., Lithopoulos, E., Reulke, R. and
Wicki, S., 2000. Design Principles of the LH Systems ADS40
Airborne Digital Sensor. /nternational Archives of Photogram-
metry and Remote Sensing, Amsterdam, The Netherlands, Vol.
XXXIII, Part B2, pp. 258-265.
Tempelmann, U., Borner, A., Chaplin, B., Hinsken, L.,
Mykhalevych, B., Miller, S., Recke, U., Reulke, R. and
Uebbing, R., 2000. Photogrammetric software for the LH Sys-
tems ADS40 Airborne Digital Sensor. International Archives of
Photogrammetry and Remote Sensing, Amsterdam, The Nether-
lands, Vol. XXXIII, Part B2, pp. 552-559.