and to enlist the aid of the Department of Photogrammetry and Surveying of 
University College London to make a valid start from the standpoint of well 
established possibilities at this magnification. 
The peculiarities of this region of operation include the need for the 
cameras to be focussed and recalibrated at a finite distance, and the 
relatively much greater focal depth over which close range cameras must work. 
On the other hand, multi-station recording becomes more feasible, and recent 
computational developments, such as the experimental bundle solution approach 
and the self calibration method (Granshaw, 1980), have become available. 
The first practical experiment in our programme was carried out at 
University College. 27 targets in a cubic array of 1m edge length were 
photographed with a single camera (Officine Galileo Special Veroplast A, 
principal distance at infinity setting 150mm, plate size 130mm x180mm) 
positioned in 8 succesive locations around the cube. The targets were white 
balls 0.7mm in diameter, set on black supports and backgrounds, and 
illuminated by the normal room lighting. Four theodolite stations on concrete 
plinths were used to perform a comprehensive survey of the targets to 
establish comparison data. The theodolite observations were also used to 
simulate the probable performance to be expected from an ideal 
photogrammetric system. The photography was on Agfa Aviphot Pan 30 (40 ASA) 
plates, measured on a Hilger and Watts stereocomparator (ium resolution, used 
as a monocomparator). 
The plate co-ordinates were treated in two different ways: 
i) Known systematic errors from a previous calibration (Scott, 1976, and 
probably not still valid ) were removed, and the resulting plate 
co-ordinates were used in a bundle-adjustment program (Granshaw, 1980) 
with no self-calibration. 
ii)The systematic errors were ascertained by the self-calibration option of 
the same program as used above. 
The experimental results are listed in Table 1. 
For both treatments, combinations of six and four photographs were 
calculated. Analysis of the variance/covariance matrix showed the following 
standard deviations of photograph co-ordinates: 
No. of photographs Calibration Standard Deviation 
6 before um 
6 self 6um 
4 before 8um 
4 self 6um 
3 self 6um 
Table 1 
The most useful features of this preliminary experiment were: 
i) The experience gained in the special difficulties of attempting this 
kind of multi-station recording with passive targets. 
ii) The limitations of cameras with low resolution and doubtful stability. 
iii)The opportunity to provide some practical data for trials of experimental 
computational schemes. 
187