he strip 
ab ars 
ww VS . 
BALLISTIC PHOTOGRAMMETRY, SCHMID 43 
From an error theoretical standpoint, it is desirable to have the optical and mechan- 
ical components of the comparator designed so that the standard error of the coordi- 
nate measuring process does not significantly devaluate the metric information inherent 
in the photographs. While there is generally agreement on the soundness of this approach, 
the opinion on the magnitude of the necessary engineering effort differs. 
This situation reflects the various experiences had with the metric quality of photo- 
graphs. In other words, there is a difference of opinion about the quantity of the residual 
error, caused by the rather complex process of taking a photograph. This error too, is 
composed of a great number of small errors. A significant component is probably con- 
tained in the variations in the image recording process, depending on the photometric 
characteristics of the object being photographed and the spectral characteristics of the 
emulsion used. In addition, a large number of residuals are present, despite careful 
camera calibration and a thorough control over the emulsion carrier and the processing 
of the photographic material. Again, the center limit theorem suggests that the sum of 
a large number of these errors has a normal distribution. 
Considerable experimental effort is still necessary before it is possible to speak quan- 
titatively about the individual error components. Results obtained from reductions of 
photogrammetric measurements in the field of ballistics make it possible however, to 
convey an idea about the typical level of overall accuracy obtainable with precision pho- 
togrammetric instruments. 
Using the Wild STK 1 Stereocomparator with 20 power magnification and a cir- 
cularly shaped measuring mark of 10 p diameter, the mean square error of a single set- 
ting on a grid plate is about = 0.7 u. The combined influence of the comparator errors 
after a thorough calibration is of similar magnitude. Thus, in the measuring procedure 
a mean square error of a single setting on a calibration grid amounts to - 1.0 y. From 
the evaluation of star photography and the measurement of signalized ground control 
points on aerial photographs, the mean square error for a single setting was found to be 
+ 1.5 y. Using three to five repetitious settings on a specific point, the mean square error 
for an individual coordinate measurement should therefore be = 1.0 pu, assuming again 
about + 0.7 u as residual error, remaining from the comparator calibration procedure. 
Actual measurements indicate this error to be = 1.5u, partly due to a bias error 
introduced by the automatic data print-out on the Wild STK 1, which works in 2 micron 
steps only. 
In the diagram c (see Fig. 3), the residuals are presented resulting from a camera 
calibration performed with a least squares adjustment, for the determination of interior 
orientation, including three coefficients for radial distortion and based on the coordinate 
measurements of 139 star images. Each measurement is the arithmetic mean of five in- 
dependent settings. The glass plate of 6 mm thickness, flat to a few fringes, was coded 
with Eastman Kodak type 103 F emulsion. The photograph was taken with a Wild Bal- 
listic Camera model No. 4 (BC-4), featuring an Astrotar lens with f — 304 mm focal 
length, stopped down to an aperture of 1:4. Catalogued star positions, being typically 
accurate to - 0.5", were used with corresponding weighting factors. The mean square 
error of a single coordinate measurement, after adjustment, was obtained as * 29 The 
distribution of the residuals is evidently close to a normal distribution. 
Diagrams a. and b. display the corresponding values obtained from the least squares 
adjustment of two spatial resections, computed with 43 and 99 ground control points, 
respectively. The corresponding mean square errors of a single plate measurement, after 
adjustment are — 2.1 y and 3 3.2 y, respectively. Again, the distributions of the residuals 
are close to a normal distribution. The photographs under consideration were taken with 
a Wild RC-7 camera, equipped with an Aviogon lens f — 10 em, with f-stop setting of 
1:5.6. Glass plates of 2.5 mm thickness coded with Gevaert emulsion No. 30 were used. 
The flying heights were approximately 4600 m and 6500 m, respectively. The influence