instances where control is minimal or marginal the results of the free net 
adjustment may prove to be superior and therefore are to be preferred. On 
the other hand, when control is abundant and strong, the free net adjustment 
can be expected to yield inferior results, particularly in situations in- 
volving weak geometry. 
3.7 SIMULATION MODE 
STAR is designed to operate in a simulation mode in order to assist 
in proper planning and optimization of potential projects. Extensive use 
of computer graphics is made to facilitate the generation of various trial 
sets of data to serve ultimately as artificial input to STAR. To start 
the process of simulation, one specifies the coordinates of points of 
interest on the object. Then, provisional starting coordinates for camera 
stations are specified along with associated principal distances, and 
dimensions of photographic formats. Following this, the computer automati- 
cally optimizes the aiming point and projects the specified set of points 
onto the format which is presented on the graphics terminal. The operator 
can then make any changes he deems necessary to the set up and can immedi- 
ately view the results of the changes. This process is iterated until the 
operator is satisfied with the set up, whereupon the image coordinates are 
assigned to a file just as if they were generated by a comparator. 
After the final camera set up has been established in the above manner, 
standard errors of image coordinates are specified, whereupon the simulator 
on command executes a least squares adjustment to generate the limiting 
standard errors of the triangulated coordinates. If these are not satis- 
factory, the operator can revise or extend the set-up and repeat the adjust- 
ment. 
Once a satisfactory set of limiting standard errors has been generated, 
the simulator executes the full-blown bundle adjustment STAR with specified 
a priori constraints on projective parameters and control points. The 
results of the error propagation are then evaluated, and appropriate revi- 
sions (if any) to the input are made. The process is continued until a 
satisfactory result is obtained or the conclusion is reached that the 
desired accuracies cannot be met under allowable or reasonable circumstances. 
Through proper exercise of the simulator one can undertake projects 
well assured that the desired objectives can be met in an efficient manner 
from the adopted operational plan. 
3.8 LEAST SQUARES RIGID BODY TRANSFORMATION 
A frequent use of close-range photogrammetry is for detection of change 
under various circumstances. However, except where one has the luxury of 
exercising a strong and abundant network of highly accurate control, succes- 
sive photogrammetric measurements are likely to be referred to significantly 
different coordinate systems. This precludes making useful direct compari- 
sons between successive sets of coordinates. To get around this difficulty, 
the STARS software package includes a least squares transformation effecting 
the combination of three translations, three rotations and a scale change 
which serves to bring any two designated sets of coordinates into best 
possible overall agreement. Through weighting, the transformation can be 
exercised in various modes. For instance, two corresponding points could 
be forced to coincide exactly, or the change in scale could be suppressed. 
The X,Y,Z residuals generated by the transformation define the discrepancies 
remaining between the two sets of coordinates and reflect the effects of 
85