248 
Since transformation by. video delay requires video in digital form, digital 
filtering and image processing techniques can easily be applied. Non-recursive 
or Finite Impulse Response (FIR) digital filtering is ideally suited to feed- 
back correlation since the time delay of such filters is independent of fre- 
quency. As a result synchronism is maintained for all components in each band 
thereby eliminating one of the problems in earlier machines. 
Digital video also offers the opportunity for compensating resolution loss 
from lens aberration and defocussing. Compensation improves both the correla- 
tion of fine detail and the resolution of orthophoto printing. It is necessary 
for compensation that aberrations be constant only over the area of a patch so 
that compensation can be separately adjusted for central, corner and interme- 
diate photo zones. 
Video delay transformation can be readily applied to the gestalt process and 
offers the following advantages: 
|. Elimination of the "ramping" of Z steps and vertical structures in the ste- 
reo model and of the consequent distortion in orthophotos and DEM's around 
such features. This resolves the first of the performance limitations set 
forth in the introduction. 
. Elimination of slope limits. 
. Elimination of sensitive analog circuitry for raster shaping and brightness 
compensation. 
4. The use of array sensors in place of vidicon image sensors and the conse- 
quent elimination of deflection yokes and analog driving circuits. 
Ww N 
The Rastar Mapping Correlator project saw the application of such a system to 
a standard analytical plotter using the single line array sensors available at 
that time (1976-78). The use of a single line sensor leads to difficulties in 
correlation and complexity in photo transport programming. These difficulties 
can now be avoided with area arrays currently available or projected for pro- 
duction in 1984, A GPM-type instrument using area array sensors could be de- 
veloped at any time and would yield hitherto unavailable speed and performance 
at essentially no increase in cost. 
The resulting ability to accommodate large vertical steps or discontinuities 
would enable stereo correlation to be used for the automatic printing of 
orthophotographs and digital terrain coordinates for large scale urban and en- 
gineering projects that seem to be the major market area for such map 
products. 
The remaining difficulty in large scale mapping arises out of tall structures 
hiding the ground and is a problem of geometry not correlation. The use of 
longer focal length cameras eases the geometry problem and allows flying at a 
greater height, usually a convenience at very large scale. Also, the smaller 
field angle improves correlator performance still further. 
Industrial robotic applications are similar to large scale mapping in that the 
surfaces of objects seen by the robot are not connected to the background nor 
to each other much the way the tops of buildings are disconnected from the 
ground. It was this similarity that led us into the robotic vision field and 
leads us to consider the application of our current technology once more to 
photogrammetry.