Nich- 
hoto- 
1ster- 
rma- 
using 
h Re- 
nsen- 
2994 
(11), 
1gsei- 
a mit 
 geo- 
CCD 
ichti- 
nfor- 
ung, 
    
FILM DEFORMATION IN NON METRIC CAMERAS UNDER WEAK GEOMETRIC CONDITIONS - AN 
UNCORRECTED DISASTER? 
Dr. Stuart Robson, Engineering Surveying Research Centre, 
City University, Northampton Square, London ECIV OHB. 
ISPRS Commission V/2. 
Abstract: 
Whilst multi-station bundle adjustments are almost universally applied to high precision photogrammetric surveys for 
engineering purposes, the use of often as few as two camera stations for less stringent architectural and archaeological 
surveys is common. The effect of in-camera deformations has been well documented in the former case, but for simpler 
geometric situations the effects and magnitudes of unmodelled distortions on the computed object space coordinates are 
less well documented. 
Self compensating effects of in-camera deformation within a strong network adjustment are illustrated with reference to 
experiments carried out using a range of film type and camera back permutations. The number of camera stations 
contributing to the network was then reduced and a range of film deformation correction techniques applied. By comparing 
results from both series of adjustments the effect of network geometry and film deformation on the computed object space 
coordinates was investigated. 
Results show that whilst root mean square object space discrepancies can vary minimally for adjustments incorporating 
differing film deformation corrections and a given network geometry, significant trends due to uncorrected film 
deformation can be present in the computed object space coordinates . 
KEY WORDS: Accuracy, Calibration, Close-range, Film, Non-metric. 
LIntroduction. 
Photogrammetry is used to measure dimensions and 
positions of objects according to some reference system 
or datum. This analysis sets out to investigate the influence 
of photographic film deformation and its correction, based 
on reseau measurement, on the object space coordinates 
derived by photogrammetric adjustment. 
The influences of un-modelled image deformations on 
derived object coordinates have been shown to be largely 
self compensating in strong network multi-station bundle 
adjustments (Fraser 1984, Wester-Ebbinghaus 1988). 
However, the majority of photogrammetric surveys 
intended for commercial purposes, for example 
architectural surveys, generally try to achieve 
photographic coverage of the object of interest using a 
minimum number of stereo pairs. This means that not only 
are any self compensating effects reduced, but also since 
there are fewer redundant measurements it is often 
difficult to determine the effects of unmodelled errors. 
This paper draws upon results produced during a 
comprehensive series of close range calibrations in which 
a variety of medium format cameras were used. Format 
sizes range from 130 by 180mm to 60 by 60mm. The 
experimental work begins with the knowledge that 
in-plane film deformation has been minimised (Robson 
1990) and that deformations occurring in the non-metric 
camera used are almost an order of magnitude greater than 
those attributed to in-plane film deformational factors. 
2. Practical Approach. 
The effectiveness of film deformation correction methods 
can be investigated by conducting a photogrammetric 
Survey of an array of targets with known X,Y,Z object 
Space coordinates. Target positions as estimated by any 
Subsequent photogrammetric adjustment based on the 
Same datum may be directly compared with the original 
target positions. Any differences in position introduced 
either by change in network configuration or by variation 
of film deformation correction can be assessed. The key 
problem associated with this method is to obtain initial 
X,Y,Z target coordinates at a greater level of precision 
than those estimated during the photogrammetric 
adjustments featuring network or film deformation 
correction changes. 
This analysis is concerned primarily with object 
coordinate deformation in single stereopairs. One method 
of deriving X,Y,Z target coordinates of suitably high 
precision is to carry out a free photogrammetric 
adjustment of the same target array based on a strong 
network incorporating many photographs. 
2.1 Test Field Design. 
A test field was designed based on several factors; the 
imaging abilities of a Hasselblad SWC camera, namely a 
resolution of at least 50 1 mmr! over the complete format 
with Technical Pan 120 film; an image measurement 
magnification of 200X and 0.5 pm rms provided by the 
modified ZKM measuring microscope at the National 
Physical Laboratory (Oldfield 1986, Zeiss 1985) and; on 
the requirement of long term durability such that the 
calibration procedure could be repeated as required. 
The final test field design consisted of 42 targets pinned 
into a concrete wall by countersunk titanium bars and 
several additional free standing targets for densification. 
The starting positions of these targets could then be 
derived by conventional survey to provide the "free 
adjustment' datum. The predominantly 2D array was then 
photographed from convergent camera positions to 
simulate a 3D array, resulting in the geometrically strong 
network detailed in figure 1. To strengthen further the 
network and enable a more precise recovery of camera 
inner orientation parameters, additional photographs were 
taken with cameras rotated by 90 degrees about their 
optical axis.