International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
  
Figure 1. SPOT Image 01 
The second test field is a panchromatic SPOT 5 image taken in 
another place in northern China in 2002, named as Image 02. 
The image contains 24000/:24000 pixels and covers 60000] 
6000 ground square kilometers. The northwest and southeast 
parts of image are urban areas and the other parts are mainly 
mountains. The ground height varies between about 10 to about 
1550 meters. The sun angle azimuth is 166.7070^, the sun angle 
elevation is 28.2098?, the principal distance of the sensor is 
1082 millimeters, the film pixel size is 3.25 microns, and the 
ground resolution is 2.5 meters. 33 well-defined control points 
were measured from 1:25,000-scale maps, 13 as directional 
points and 20 as check points. Figure 2 illustrates the coverage 
area of the image with the ground points distribution. Triangles 
represent directional points and circles represent check points. 
  
Figure 2. SPOT Image 02 
4.20 Results and Evaluation 
The least square estimator, the stein estimator, the ordinary 
ridge estimator with Horel-Kennad and Lawless-Wang 
approaches, the generalized ridge estimate and the combined 
ridge-stein estimator are used to compute exterior elements for 
these two SPOT scenes. LS refers to the least square estimator, 
Stein refers to the stein estimator, HK refers to the ordinary 
ridge estimator with Horel-Kennad approach, LW refers to the 
ordinary ridge estimator with Lawless-Wang approach, GR 
refers to the generalized ridge estimator and CRS refers to the 
combined ridge-stein estimator. 
Root mean square errors (RMSE) in the image photo 
coordinates and the ground space coordinates are taken to 
evaluate precision of different orientation methods. vox, voy, 
Vox, Voy and Voz represent directional points' RMSE in the 
image photo coordinates xL1y and object space coordinates X, Y 
and Z respectively. And vex, vey, Vex, Vey and Vez have the 
same meaning for check points. 
For test data Image 01 Table 1 and Table 2 show RMSE at 
directional points and check points respectively. 
Unit: pixel(p) and meter(m) 
  
  
  
  
  
Method vox/p voy/p Vox/m Voy/m | Voz/m 
LS 61.80 20.10 715.67 “21782 30.16 
Stein 1.07 1.20 8.27 10.88 735 
HK 1.05 1.2] 8.18 10.82 7.34 
LW 51.88 10.92 66022 11273 25.53 
GR 0.98 1.01 7.66 10.07 7,33 
CRS 0.82 0.91 5.67 8.51 7.32 
Table 1. Test results at directional points for Image 01 
Unit: pixel(p) and meter(m) 
Method vex/p vcy/p Vcx/m Vcy/m | Vcz/m 
LS 72.37 26.68 02234. 125.62 22.30 
Stein 1.54 1.67 13.92 13.14 7.83 
HK 1.49 1.66 13.70 14.89 — 7.85 
LW 61.75 12.13 675.49 111,88 20.71 
GR 1.30 1.41 13.48 14.43 7.84 
CRS 1.23 1.38 13.16 13.73 1-17.58 
  
Table 2. Test results at check points for Image 01 
For test data Image 02 Table 3 and Table 4 show RMSE at 
directional points and check points respectively. 
Unit: pixel(p) and meter(m) 
  
  
Method vox/p voy/p Vox/m Voy/m | Voz/m . 
LS 123 1.26 4.11 4.66 5.40 
Stein 0.46 0.27 2.90 2.26 1:99 
HK 0.44 0.26 2.78 2.25 1.99 
LW 105.81 134.40 710.7 745.3 123.32 
GR 0.41 0.23 2.69 2.22 1.98 
CRS 0.22 0.20 2.02 1.80 1.54 
  
Table 3. Test result at directional points for Image 02 
730 
Inter: 
  
  
LS 
Stein 
HK 
LW 
GR 
CRS 
  
Tab 
Table 
estim: 
SPOT 
precis 
contre 
estime 
with ! 
estim: 
pixels 
estim: 
ordinz 
latter 
estimz 
appro: 
There 
the m 
pushb 
The si 
linear 
condit 
optim: 
estima 
and in 
that th 
the ri 
reliabi 
perfec 
Gui, € 
estima 
Geode 
Guo, | 
genera 
orienta 
In: 77 
Sensin, 
XXXI 
Gupta, 
Trans. 
Huang 
Applic,