Paper for Commission III (Aerial Trian lation) Of Ninth 
INTERNATIONAL CONGRESS 1960 
J. A. Weightman 
STEREOBLOCK ADJUSTMENT 
1. General 
It is fundamental to all survey computation that it is more 
accurate to adjust large sections of a network at a time, rather than 
build up piecemeal from very small adjusted units, Thus, in ground 
triangulation, it is preferable to adjust polygon by polygon, rather 
than triangle by triangle,* and better still to adjust the whole 
network as a unit, 
Ihe method of variation of coordinates starts with approximate 
coordinates (either in the plane or in Space) of all points of the 
network, derived in any convenient manner, and expresses small 
changes in the observed quantities as linear functions of small 
changes in the relevant coordinates, These functions are then 
equated to the discrepancies between the quantities as observed 
and their values as derived from the approximate coordinates, A 
least squares solution is computed from these observation equations 
to give those changes to all the coordinates which will most nearly 
eliminate these discrepancies, 
This procedure is in constant use for rigorously adjusting quite 
large blocks of ground triangulation, and is of course the basis of 
the method originally formulated by Professor Church for adjusting 
blocks of aerial triangulation, A disadvantage of the application 
to a block of photography of any size appears to be that the 
number of variables involved becomes quite unmanageable, 
The object of the present paper is to propose a compromise to 
bridge the gap between adjustment in large simultaneous blocks, 
which may be uneconomic, and adjustment by stereograms, which may 
not attain the full accuracy possible, 
  
* Compensating errors in any particular triangle are 
shown up by the side conditions, and so eliminated, whereas 
if one merely adjusted by a chain of simple triangles, these 
errors would pass "undigested" right through to the final 
framework, even after this framework had been stretched to 
fit on to any ground control available, 
x A typical ground triangulation scheme of LO new points 
would lead to normal equations in 80 variables; an equivalent 
aerial scheme of 15 strips of 20 photographs, with 6 minor 
controls per overlap and 20 ground control points for the 
whole block, would lead to normal equations in no less than 
2820 variables, 
If one reckons 100 man-days hand computing, or 40 minutes 
electronic computer time, to solve the first set of normal 
equations, by Gauss-Doolittle or Cholesky, and as the time 
to invert a matrix varies as the cube of the number of variables, 
one might at this rate expect the computation of the serial 
triangulation block to take some 10,000 man-years by hand 
computing, or some 28,000 hours of electronic computer time, 
A formidable period! 
  
  
I unse