Page A1

Appendix A - Identifying the Kurt Franz Camera

Appendix A - Identifying Kurt Franz’s Camera

The confident identification of existing buildings imaged on both ground and aerial photos required triangulation to pinpoint the position from which the terrestrial pictures were taken and the angles to the various buildings visible from ground and air stations. To do this required a knowledge of the type of handheld camera used, its nominal focal length, its field of view (Fov), and the format of the film used. Several makes of German cameras produced in the era between 1930 and 1943 were considered. The cameras surveyed are shown in the table below.

Camera Model/Year Lens FoV Film Format
Leica/1930 Hector 50mm f/25 37.6 degrees 35mm film - 24 x 35mm
Voightlender Bessa/1936. Folding Belows 105mm f/2.8 46.4 degrees and 24.2 degrees 120 film with two options: 90 x 60mm or 45 x 60 mm
Kodak Retinette/1939 45mm f/2.8 40.7 degrees 35mm film - 24 x 35mm
Minox/1938-43 15mm 40.1 degrees 8 x 11mm
Table A1 - Candidate Cameras Evaluated

The Leica’s inclusion afforded a generic type for all 50mm focal length, 35 mm film cameras. The other cameras encompassed a variety of formats. The Minox was produced for the German army in WWII. The Voightlander and the Retinette covered commonly produced cameras available to the German public. Of these camera's, analysis showed that only one could have been the instrucment used by Kurt Franz: the Voightlander Bessa, a bellows type camera with a relatively long nominal focal length of 105mm. The Voightlander had two modes of operation ; one could expose eight 90 by 60 millimeter pictures, or sixteen 45 by 60 millimeter frames. Analysis revealed that the Franz photographs were exposed in the smaller format.

The determination between the candidate cameras depended on matching the geometric characteristics embedded in the ground photos to the ground features which could also be identified on aerial photographs. The procedures was as follows. The angular spacing those features were first determined. This procedure is illustrated in Figure A1, one of the pictures so used. In the photograph, the various features were measured from an origin at the left. The ratio of the overall image width to the measurements (1) through (9) could then be used to calculate the respective angle for each camera system by:

Angle(n) = measure(n)/W x FoV x SF ,

Where: W = width of the print being measured,
FoV = camera field of view,
SF = scale factor of print to original negative.

The measurments made were then plotted on a transparent meduim. The height of one or more features was also determined by shadow length methods, and this datum used to confirm the locus of the ground camera’s exposure station.

The angles measured in Figure A1 were drawn on a transparency (Figure A2) and then were slid over an aerial photograph and moved until a good visual fit of the various lines was made with picture elements originally measured on the ground photo but as seen from the air.

The the shadow height measurments provided an independent check on the visual fits obtained using the transparent overlays. An overlay was generated for each of the candidate cameras. Some of these plots seemed to fit the aerial photogaphs, but when the shadow heights information was used to compute the exposure station, a large discrepancy occurred. Only one camera - the Voighlander Bessa - provided consitent fits in which both the overlay and the camera station measurments agreed.

Shadow Height Measurements

The height of several features imaged on the ground shots was determined as follows. First, the sun elevation () and azimuth () was established at the time the aerial photography was exposed. The azimuth was measured directly on the aerial frame after it was accurately aligned with true north. Shadows cast by buildings and trees were then measured, and an average taken. It was found that on the frames of the GX120 mission, = 90 degrees. That angle, in conjunction with the geographic coordinates of the camp and the time of year, was then used to compute the local sun time (t) and elevation angle. The results were: t =0634 AM, and = 24 50’.

The last step was to measure on the aerial photos the length of the shadows cast and then to compute the heights of the features casting them. This is given by elementary trigonometry by:

Height = SL * Tangent()
. Where: SL = shadow length, measured on the aerial photos.

Other Height/Ranging Measurements

The distance of objects from the ground camera whose height could be estimated with some confidence were also computed as a further check. Estimates relied on internal clues contained in the images. Figure A3, shows a well and two of the Jewish workers. It has been enlarged from a Kurt Franz picture. A fairly good assumption is that the height of the figure on the left is about 1.7 or 1.8 meters. Given this assumption, the scaled height of the structure used to support the windlass shaft is 1 meter. The well shown here is also visible in the previous photograph ( Figure A1 ) as element number 6. To determine the distance from the camera station, one must knows the height of the feature in object space; the respective measurement in the image space; and the focal length of the camera. The distance to the object (D(o) is then given to a first approximation by the following equation (see Figure A4:)

D(o) = {FL * H(o)}/H(i) ,

FL = Nominal focal length of the camera
H(o) = Estimated height of the object
Where: H(i) = Measured image height of the object
SF = Scale factor to compensate for enlargements

More precise photogrammetric procedures were unmerited. It would have entailed control points in the scene, camera calibration, and a knowledge of camera tilt, roll, and pitch angles. These data were obviously not available.

Illustration of Data Fit

The image in Figure A5 demonstrates the fit of the angular measurements plotted on transparent overlay materials fitted to the aerial photograph. The overlay was rotated and translated until each line intercepted the corresponding conjugate image feature as closely as possible. As can be seen in the illustration, the fit was quite close. The next step was to see that the range measurements agreed with the distance from the camera station through the visual fit procedures. This step provided a cross check. If the wrong camera had been assumed, the range measurements and angular orientation of the camera cone angle could not have been reconciled. As it turns out, the distance of the camera station established by the visual fitting, to the position indicated by the intersection of the three range loci (shown in the next figure) is only about 7.5 meters. This figure is a very satisfactory error considering the tremendous unknowns and many sources of errors in the methodology. The same procedures were used in two other cases where ground photos of the living camp were available. In addition, all the candidate camera systems were evaluated. Only the Voightlander Bessa’s characteristics yielded a good fit of the angular data to the aerial photography.


Last modified: November, 2004
Copyright © 2003 Charles A. Bay. All rights reserved.
Technical/administrative contact: