154 
repeated in the next test where height differences and distances are used in 
the combined adjustment. Table 4 shows the effect of the combined adjustment 
on the redundancy numbers. There is an additional improvement in rOx) (about 
25%) and no change in r(y). However, the improvement in the effect on object 4. 
coordinates is substantial especially when the error is in x coordinate 
(case A). In this case the object coordinates are almost unchanged due to the 
error. When the error is in y (case B), the resulting error in Z is almost 
  
  
eliminated while the errors in X and Y are reduced slightly. Ta 
T us 
It is now clear that the combined photogrammetric and terrestrial adjustment in 
has a great advantage in improving the reliability both internal and external. wi 
All that is needed is the measurement of distances between points (two te 
distances to each point) on the perimeter of the block where the reliability be 
is originally the lowest. Height differences are not needed for cases where po 
the ratio between variation in terrain elevations and camera station height pe 
is large enough to cause correlation between planimetric and height ar 
coordinates such as in close range photogrammetry. However, in cases of in 
nearly flat terrain, height differences will help at least in improving the is 
external reliability. 
(b 
Effect of Additional Constraints on Some Systematic Errors 
Th 
Since systematic errors are much smaller than most gross errors and affect all si 
the points in the block, it is expected that the influence of the combined an 
adjustment will be very different on the two types of error. In the case of ra 
systematic errors it will probably depend more on the source of error and the re 
distribution of the terrestrial observations. Since many factors are needed no 
to be studied here, only the simulated block is used ín the following tests. pa 
CO 
(a) Image Coordinates Contain Radial Lens Distortion: ef 
er 
an 
A generated lens distortion data, using the Wild Aviogon lens distortion 
curve, have been added to the simulated image coordinates. The following di 
parameters are studied: 
a. type of terrestrial observation Th 
b. number and distribution of terrestrial observations ad 
c. number of control points 
£o 
Various tests have been carried out with the results displayed in table 6 
(tests 1-8). The different distance distributions are shown in figure 4. The 
height differences are at the perimeter of the block. Control point de 
distributions are also shown in that figure. Analyzing these tests, the 1. 
following comments can be made: 
1. The overall effect on the residuals is negligible. The standard 
error of unit weight has not changed while the residuals at 
individual points have changed slightly up or down, 
2. When no terrestrial data have existed, the control point distribution is 
critical (compare object coordinate error in tests 1 and 2) while 
additional control points do not improve the results significantly in the 
case of combined adjustment (compare cases 3 and 4). 2. 
Comparing test 1, where 20 planimetric and 34 vertical control points 
have been employed without additional constraints, with test 8, where 8 
planimetric and 14 vertical control points have been used with terrestrial 
observations, it is clear that the terrestrial data not only replace many 
of the control points but also improve the accuracy. 
3. The optimum distance distribution is 28 perimeter distances (test 6). 
These distances do not form a closed polygon around the block like in test Al 
te 
3 but have few gaps which have not affected the accuracy but, ou the other