170
technology has given rise to automated techniques for
performing this task.
The goal of this investigation is to develop and test a
system which can automatically generate a digital terrain
model using digitized aerial photography and a general
purpose computer. This system utilizes a scanning micro
densitometer to convert aerial photography to a form which
can be "understood" by the computer. The densitometer
scans the photography in a very fine grid pattern and
records numerical values indicating the gray-shade at each
point of the grid. In order to establish photogrammetric
relationships this arbitrary scanner coordinate system must
be transformed to the fiducial axis system. This requires
locating the fiducial marks in scanner coordinates and
transforming them to the calibrated values in the photo
graphic XY system. When the parameters of the transforma
tion are determined, photo coordinates can be readily
computed for other points in the scanned image. Basic to
any photogrammetric system is the ability to locate conju
gate imagery on a pair of overlapping photos. In the
automated mapping system this process is performed by a
technique known as digital image correlation. In this
method, two arrays of image densities are compared
numerically to indicate the degree of similarity of the
images. Pass points and ground control points are located
by this method allowing a numerical relative and absolute
orientation, or a simultaneous bundle adjustment to be
performed. Once the orientation parameters are known a
regular grid of digital terrain model points can be corre
lated and absolute XYZ positions determined. This data is
then input to a contour mapping program to generate the
topographic map.
Two sets of photography have been used to test the system.
The first set, taken of an area near LaCrosse, Wisconsin,
was used to develop the DTM generation software. Using
these photos, a digital terrain model was generated auto
matically and compared with a DTM read manually in a Kern
PG-2 stereoplotter. The second set, taken over Mazomanie,
Wisconsin, was used to test ground control panel recogni
tion and scanner calibration. These photos contained
specially designed control targets which were automatically
located and converted to photo coordinates. These coor
dinates were then compared to values measured with an H.
Dell Foster monocomparator to indicate the accuracy of the
automatically located panels. Bundle adjustments were
performed using the monocomparator and automatically
derived photo coordinates and results compared.
SYSTEM HARDWARE
The instrument used to convert the photography to digital
form was an Optronics 1700 scanning microdensitometer.
This scanner employs a rotating drum and lead screw to
uniformly move the photo past the scanning optics. At
each point along the scanning grid a narrow beam of light
is transmitted through the film. The transmitted light
intensity is compared with the incident intensity to obtain
a digital value which indicates the relative darkness of