It is therefore possible to calculate the photo coordinates of each ortho-
photo pixel as it would appear on the aerial photograph (Fig. 6). It is
only necessary to sample the photograph at each calculated location and then
to play back those digitized pixel values in a regular grid pattern to
generate the orthophoto.
RESAMPLING
The locus of the photo coordinates of a row of orthophoto pixels will be a
non-orthogonal line except when the elevation is the same for all pixels in
the row. If the aerial photograph were to be scanned in a rectilinear
fashion, there is little likelihood that any of the sampled locations would
correspond with the desired photo coordinates. Therefore, some type of
resampling of the pixel data would have to take place.
Incorrect pixel opacity values are a drawback of resampling. The calcula-
tions involved can be time-consuming, too, unless performed by a high-speed
computer, while the great amount of data which must be stored and manipulated
requires that the computer also have a large capacity. If the same piece of
equipment is to be used for both scanning and playback functions, then all
of the data extracted from the input photo must be stored between operations.
Manipulation of this data subsequent to its being digitized can be expensive,
as noted above. Therefore, pre-sorting of the data by an appropriate
sampling method can have distinct economic advantages.
NON-ORTHOGONAL SCANNING
One approach to digital orthophoto production would require that each
pixel of the final orthophoto be sampled, digitized, and stored in its
proper serial sequence by addressing, in turn, successive spatially-inde-
pendent aerial photo coordinates. A normal, rectilinear playback of the
stored data would complete the process. A non-orthogonal scanning capability
must therefore be added to the photodigitizer used for this purpose.
The gross movement required to scan a 9-inch frame of aerial film could be
provided by an X-Y stage assembly. But such a stage is only suitable for
scanning in straight, orthogonal lines at a uniform velocity. Subtracting
such an orthogonal scan from the non-orthogonal locus of desired photo
sampling locations leaves a residual x- and y- displacement for each
location (Fig. 7).
A piece of hardware suitable for accommodating these displacements is an
image dissector tube (Fig. 8). This is a multi-stage photomultiplier tube
which has the capability of electronically deflecting the photoelectrons
generated by an optical image focused upon its photocathode. This allows
the "electron image" to be scanned past a fixed aperture located upstream
from the multiplier section of the tube. The square aperture can be
"positioned" in X and Y at any one of 4096 locations per axis, with a
spatial resolution of 1/20 of the aperture size and an accuracy of +/- 1%.
The effective aperture size at the film plane (pixel size) depends upon
the overall magnification of the imaging system, but should be variable in
size from 10 to 50um per side. The image dissector tube operational cycle is
no more than 50usec per pixel for a maximum sampling rate of 20,000 pixels
per second.
64
st
(
\