observations. 
nd two orbits 
  
TSize of 
pixel 
1150 (m) 
  
  
  
  
| errors in the 
sition, in the 
s. Ten of the 
errors". Their 
1), that is, the 
values. There 
p 2). 
a, are applied 
All ten "large 
y the Sequent 
Robust method 
ordinates. The 
rrors", greatly 
jiaarda method 
  
Baarda 
  
10/10 
  
10/10 
  
1/1 
  
  
  
  
  
luring the first 
or the Baarda 
tment. In this 
the ten "large 
; method finds 
thod has found 
ee gross errors 
ults show, we 
and the Baarda 
method, or the Sequent and the Robust method, in the bundle 
block adjustment. 
5.2 Variance estimation 
The simulated data contain five kinds of observation: two 
groups of image coordinates (two cameras), two groups of 
position orientation observations, and one group of object 
coordinates of the control points. We can simultaneously 
estimate five a priori variances of the observations during the 
adjustment. Only the priori variances of the image coordintes 
are estimated, or the a priori variances of image coordinates and 
the position observations are calculated. We have given all 
  
  
  
  
  
  
  
show that the estimated offsets and drifts are in good agreement 
with their true values (Tables 6 and 7) 
6 . RESULTS FOR CLEMENTINE IMAGE DATA 
In 1994 approximately 600 000 images of the Moon were 
obtained from the UVVIS CCD-camera onboard the Clementine 
Mission (Table 8). 
Stereo image sequences were taken in two areas near Mare 
Orientale. In this particular study area the ground pixel sizes of 
the images varied from 125 to 250 m. We have derived from 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
modules, with different values of variances. The results of two C1(x0) C2(y0) C(x) Number of 
examples are listed in Table 4 (example I) and Table 5 (example (m) (m) (deg) errors in 
ID). Three different initial values are used in each test (initial 1, other nav. 
initial 2, and initial 3). All results show that the estimated a true 4000 2000 0.5 : 
priori variances are identical with their true values, and they are values 
not dependent upon their initial values (Table 4 and Table 5). times 1 2457 239 0.21 
+/-560*) _|+/-89 +/-0.07 3 3er) 
Values Image Position — | Orientation |Control times 2 4047 1739 0.59 
points data data points 44-29 44-2271 +/-0.04 0 
(mm) Hm) (deu nn results [4047 1632 0.59 
true 0.02 100 0.01 200 17-30 4-256 +/-0.04 0 
0.005 500 0.05 
initial 1 1.0 200 0.01 600 Table 6: Elimination of systematic errors for example 1 
1.0 1500 0.15 *) standard deviation. 
estimated 1 10.0196 99.4 0.095 196.5 **). "3" means that besides the offsets Cj(,0), Co(y0) and 
0.0051 450.8 0.055 C(x)» three offsets in the other navigation data are used as 
initial 2 0.01 200 0.01 600 unknowns in the second adjustment. 
0.01 1500 0.15 
estimated 2 |0.0196 88.8 0.011 195.9 
0.0051 454.6 0.054 C1(x0) C2(y0) Cy( x) Number of 
initial 3 0.0001 200 0.01 600 dy (x0) d5(y9) ax) errors in 
0.0001 1500 0.15 (m,/km) ]|(m,/km) |(deg, /km) other nav. 
estimated 3 |0.0196 88.8 0.011 195.9 true 2000 1000 1.0 - 
0.0051 454.6 0.056 values 2 5 0.01 
times 1 705 557 1.16 
Table 4: A priori variances for the example I +/-329 +/-246 +/-0.05 
171 3.42 0.0208 4 
+/-1.35 +/-0.79 +/-0.0002 
Values Image Position Orientation |Control times 2 908 1500 1.22 
points data data points +/-300 +/-708 +/-0.09 
(mm) (m) (deg) (m) 4.25 4.12 0.0197 0 
true 0.1 2000 0.2 1000 +/-1.60 +/-2.41 +/-0.0003 
0.1 2000 0.2 results 2058 646 1.09 
initial 1 1.0 10000 1.0 5000 +/-64 +/-354 +/-0.07 
1.0 10000 1.0 1,95 4.94 0.0200 0 
estimated 1 [0.095 2204.8 [0.235 958.9 +/-0.24  |+/-0.97  |+/-0.0000 
0.101 1826.3 0.23 
initial 2 0.01 10000 1.0 5000 Table 7: Elimination of systematic errors for example 2 
0.01 10000 1.0 
estimated 2 [0.095 2187.8 0.238 998.9 
0.101 1888.1 0.228 : 
initial 3 |0.0001 [10000 1.0 5000 Clementine 
0.0001 |10000 1.0 mum UV VIS 
estimated 3 0.095 2167.0 0.236 991.1 CCD chip size 384 x 288 
0.101 1918.3 0.235 focal length (mm) 90.000 
pixel scale (pix/mm) 43.478 
Table 5: A priori variances for the example II 
5.3 Elimination of systematic errors 
The simulated navigation data contain three offset parameters 
(example 1), or three offset and three drift parameters (example 
2). Using our methods described in section 4.3, we have 
determined the systematic errors in the adjustment. All results 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
  
  
  
  
Table 8: Clementine UV VIS camera 
these a local control point network, and two Digital Terrain 
Models (Oberst et al., 1996). We consider only the bundle block 
adjustment for the two areas, and analyse the results. The image 
data for the first area (Test 1) are taken from orbit 332. The 
1007