)- 
c). 
ri- 
anal 
nal 
che 
ust- 
ried 
age 
st- 
        
then results are obtained that compare favorably 
with simultaneous planimetric adjustments. The se- 
quential method is very simple, fast and numerically 
stable. The simultaneous methods are expensive, 
slow and numerically unstable. 
Figure 5 illustrates the results achieved in 
sequential adjustment of actual radar blocks 
PRORADAM and West Virginia to sets of ground con- 
trol points. It appears that with a density of 2 
ground control of about 15 points per 100 000 km , 
point accuracies achieved are about X150 m. 
  
  
  
  
  
E 3.0 I I I 
S 5 sl. A PRORADAM 
7 e 
$ ® W. VIRGINIA 
S 2.01 
QO 
wi 
I 
U 1.5} 
z 
ec 
9 tot 
= 
LL 
pe 
Z 0.5} 
O 
a 
wv 0 i | | 
S9 5 10 15 20 
CONTROL POINTS PER 100,000 km? 
7.3 Mosaicking: 
Mosaicking of radar strips can be carried out 
using the results of the numerical adjustment as 
a base map. However, the individual radar strips 
cannot be rectified, except for elimination of 
scale differences along- and across-track.  There- 
fore, one must expect that mosaicking itself con- 
tributes a sizeable mapping error, even if the 
results of the numerical adjustment were entirely 
error-free. 
In the case of the block West Virginia, check 
and ground control points were scaled off the 
mosaics at scale 1:250 000 and off the topographic 
maps at scale 1:24 000, using in both cases the 
grids printed on the map sheets. It was found 
that both groups of points had the same errors of 
approximately £200 m. This shows that the mosaick- 
ing of largely unrectified radar images added an 
error that obscures the errors of the numerical 
adjustment (£130m). 
Other methods of mosaicking have been and are 
in use. A method applied in the initial phase of 
Brasil's project RADAM is based on complete 
SHORAN tracking of all mapping flights. An ordered 
set of image points is then transformed onto a 
base using the measured aircraft positions and 
image ground ranges. The images are laid out on 
this base. This results in mosaics with r.m.s. 
errors of about £300 m (v. Roessel and de Godoy, 
1974). 
In most of the enormous RADAM project of 
Brazil (9 million km ), however, not all of the 
mapping flights were SHORAN controlled: in fact, 
none of the production flights were controlled, 
and only a small number of transverse cross-lines 
(tie-lines) were. Mosaics were laid out using 
the SHORAN controlled tie-lines as a geometric 
reference. This method of mosaicking was estimated 
by v. Roessel and de Godoy (1974) to lead to errors 
of the order of magnitude of £700 m. 
It is thus obvious that mosaics based on a 
numerical  planimetric adjustment are more accu- 
rate than those resulting from other methods. It 
can even be expected that the numerical approach 
is the least expensive one (Leberl, Jensen et al., 
1976). 
7.4  Three-Dimensional Adjustment of Radar Blocks: 
  
The only 3-D block adjustment with actual 
radar imagery so far has been reported by DBA- 
Systems (1974). In this adjustment the flight 
data are described by spline functions of time t. 
The coefficients of the splines are solved in the 
simultaneous adjustment, based on the linearized 
projection equations (6), (7). Also the measure- 
ments of slant range and time are being calibrated 
in the simultaneous adjustment by postulating an 
error behaviour according to polynomials: 
- 2 2 3 
t ve = t. + ar + a,r + a,r 
(21) 
in 2 
y + = To + by + b,r + bar 
t, r are observations, ve? v_ are corrections, 
to r are approximations, and a,, . . . b 
unknown. The program employed i DBA-SysBens is 
thus similarly general as are the well-known 
photogrammetric bundle adjustment programmes with 
additional parameters. 
  
  
  
CONFIGURATION CONTROL ACCURACY REMARKS 
per , (Meters) 
100 km 
along across height 
lc lc lo 
Single Image " 09.5 519 26 197  LIutetfer 
rometer 
Stereo Pair 1.2 26.8 21.9 . 16.7 Opp. Side 
Stereo Pair 1.2 29,5 25.6 19.7 Same Side 
: Opp. and 
Triplet 142 18.9 11.4 . 14.5 Same Side 
  
  
  
Table 5: Mapping accuracy achieved using single radar ima- 
ge with interferometer and also stereo- and triplet inter- 
sections (From DBA-Systems, 1974) 
Table 5 reviews the results obtained by DBA 
(1974) with a block of three radar images. Use 
was made of interferometer measurements in the 
single-image approach and of HIRAN tracking of the 
aireraft. 
The very interesting result of Table 5 is that 
even same-side stereo provides higher accuracy than 
use of the interferometer in conjunction with a 
single image. The study by DBA-Systems (1974) al- 
so indicated that inclusion of interferometer data 
in the stereo or triplet intersection does not im- 
prove the results. 
8. CONCLUSIONS, RECOMMENDATIONS 
8.1 Review of Present Status: 
  
The period since 1972 has, for radargrammetry, 
been marked by the declassifigation of military 
know-how (studies and 3 x 3 m^ resolution imagery), 
extensive operational reconnaissance type mapping