shape of the object.
TABLE I Internal precision results for the multistation configuration
illustrated in Fig. 1.
case Control Target Internal precision
Angormation number X Vy Z
A No control 1 0.175 0.155 0.175
information 2 0.181 0.153 0.172
3 0.172 0.153 0.181
4 0.179 0.152 0.179
B 24 corner points 1 (0.042) (0.042) (0.042)
used as control 2 0.180 0.151 0.169
3 0.169 0.151 0.180
4 0.179 0.152 0.179
C camera exterior 1 0.164 0.146 0.164
orientation parameters 2 0.176 0.147 0.165
held fixed 3 0.165 0.147 0.176
4 0.176 0.149 0.176
In case B the 24 corner points were used as control (although the internal
precision figures still refer to the network of targets as a whole), producing
only a marginal increase in precision. Even in case C, where the six exterior
Orientation parameters of each camera station were held fixed at their true
values, only a 4$ increase in precision is achieved over case A, illustrating
the inherent strength of a good multistation relative orientation. Also note
the relative homegeneity of the figures in the three co-ordinate directions,
which is not a general characteristic of conventional stereophotography.
Conclusion
High precision and accuracy are important aspects of many close range
photogrammetric solutions but their means of attainment are, on occasions,
biased by the vast photogrammetric experience in mapping from aerial photo-
graphs. This expertise has lead to the mapping application being considered
the photogrammetric "norm" whereas scientifically it is, of course, a special
case in three important respects. Firstly, aerial photography uses a fixed
infinity focus metric camera. Secondly, the geometry of aerial configurations
and the shape of the object (the ground) are highly specialised. Finally,
stereoscopic measurement of detail is commonplace and must be carefully
considered in the planning of such photography.
Only by close range photogrammetrists viewing the camera purely as an angle
recording instrument, and differentiating the measurement function from the
correlation function, can a rational decision on suitable approaches be made.
Stereopair geometry will still be relevant to many applications, but the
multistation method is worthy of consideration in several high precision
areas, such as the co-ordination of industrial and engineering structures,
with its advantage of high, homogeneous precision without control measurements,
the ability to compensate systematic errors, and the ready detection of gross
errors. Finally the question of what the precision estimates are relative to
242