International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 Internat 
3.3 Ground control 3.4 Image processing 
During a field survey, ground control points were measured in The images were processed using the IMAGINE OrthoBASE 
the Mam Tor area using two geodetic quality Differential GPS Pro 8.6 software package (ERDAS LLC, 1991-2002). An T | 
single frequency receivers. One receiver served as base station, independent module performing the self-calibrating bundle 
while the other was used to collect GPS data at the selected adjustment was used for estimation of the interior parameters. 
ground control points. From a nearby active GPS station (part This software (GAP, General Adjustment Program) was written 
of the National GPS Network, established by Ordnance and developed by Chandler & Clark (1992). 
Survey). additional GPS data was obtained and used for the 
post-processing. Post-processing revealed the positions of the 3.5 Results and accuracy 
measured points in ETRS89 (European Terrestrial Reference 
System 1989) coordinates, with an precision of approximately The self-calibration adjustment was performed successfully for 
1cm. The processing was performed by using SKI-Pro software all sets of photographs, with acceptable residuals of the control 
(Leica/Geosystems). Finally, the coordinates were transformed points. The self-calibration also provided an estimated object 
to national grid coordinates (OSGB36 datum) using the online precision, based on the stochastic properties of the control 
coordinate transformation provided by Ordnance Survey points. The low height accuracy from the 1953 photographs can 
(www.gps.gov.uk). be attributed to the low base-height ratio, and hence weak 
geometry. À summary of the self-calibration performances is am 
given in Table 3. Figure ! 
1953 images 
From the scanned imagery high resolution DEMs were 
Focal length (mm) 547 extracted. In addition to the estimated object precision from the source 
Base length (m) 750) self-calibration, the DEM quality can be specified by a global digital 
Flying height (m) 6140 Root Mean Square Error (RMSE), indicating the difference Suena 
B/H ratio 173.2 between the height values of the control points and their may le: 
(X) (Y) (Z) corresponding DEM values (see Table 4). object 
Residuals ground control (m) 0.57 0.41 0.09 sapie : 
Residuals photo control (u) 51.99 33.26 In addition to a poor base-height ratio, there are two other Th 
Precision of object points (m) 0.48 0.99 4.26 possible causes of errors. Firstly, discontinuities in the terrain ased'o 
are downgraded by interpolation (Petrie. 1990). Another error Chang 
1971 images : dea 
Epoch Resolution RMSE 
Focal length (mm) 304 1953 4.0m 3.90m mens 
Base length (m) 580 1971 2.0m 1.31m i — 
Flying height (m) 2240 1973 0.76m 1.04m A DEN 
B/H ratio 1/3.9 1990 2.0m 0.54m s T 
(X) (Y) (Z) Table 4. Summary of DEM extractions. SPDrOA 
Residuals ground control (m) 0.45 0.54 019 The h 
Residuals photo control (u) 30.22 35.46 displace 
Precision of object points (m) 0.21 0.25 0.91 (Figure 
the 
1973 images the kr 
epochs; 
Focal length (mm) 153 e M 
Base length (m) 280 precisic 
Flying height (m) 950 sample 
B/H ratio 1/3.4 x-direct 
(X) (Y) (Z) the lanc 
Residuals ground control (m) 0.085 0.12 0.17 confide 
Residuals photo control (u) 19.66 18.59 
Precision of object points (m) 0.075 0.088 0.20 
1990 images 
Focal length (mm) 153 
Base length (m) 1010 
Flying height (m) 2120 
B/H ratio 1/2.1 
(X) (Y) (Z) 
Residuals ground control (m) 0.025 0.021 0.013 
Residuals photo control (u) 15.44 16.03 
Precision of object points (m) 0.18 0.14 0.40 Figure 5. 3-D views of the centre part of the landslide: an 
orthophotograph (1990) and a DEM -of-difference 
  
  
  
Table 3. Summary of the self-calibration performances. (1990-1973) draped over a DEM. 
478