for ail portions of the photograph due to water areas; clouds, haze, 
or imperfections of the film. Therefore, a pre-selected array of 
image coordinates may not always be measured. It would be advantageous 
to establish some dynamic criterion, amenable to on-line computations, 
which would define the most effective location for an image measure- 
ment of a pass point from a list of candidate coordinate locations 
as the measurements were being made. Since the image coordinates will 
be incorporated into the aerotriangulation least squares solution, the 
selection of observations in the form of image coordinates becomes a 
problem of statistical experimental design. 
The statistical design of experiments has been given much attention 
in the literature. Authors (4,6) have addressed the problem of 
determining optimum observations for standard regression analysis 
Eroblens Box and Lucas (2) have suggested that observations be used 
which make the determinant of the inverted normal equations as small 
as possible, or conversely, the determinant of the matrix of normal 
equations as large as possible. Some statisticians have provided 
criteria for judging the "goodness" of an observation based on 
statistical information theory.  Lindley (10) using information 
theory, derived theorems which give the amount of information provided 
by an experiment.  Neyman (13) also used information theory to 
establish a criterion for optimum station selection for geodetic 
Surveys. Using some of the theory and principles of statistical