Aerial Photo - Orthophoto Primer

1
Aerial Photo - Orthophoto Primer

• Bill Befort, Remote Sensing Coordinator
• Resource Assessment Unit, DNR Forestry
• Grand Rapids MN

2
Photo Projects Map Large Areas in Great Detail
3
But Photos Arent Maps --
A map is an orthographic view, and shows every
object as if from directly above it . . .
whereas even a perfectly vertical airphoto is a
perspective view from a central point.
4
. . . maps have uniform scale, projection,
orientation, and symbology
. . . whereas photos do not.
5
Vertical airphotos have maplike qualities

• --just dont count on it

6
. . . but theyre seldom perfect. They exhibit
!
Tip (n)
Tilt (S)
Swing (6)
and sometimes even (ugh) dIsToRtIoN
7
And they all exhibitTopographic Displacement
Map
Vertical photo (same nominal scale as map)
Oblique view
8
. . . which is bad, and good.
Topographic displacement (the effect of
perspective) is the least tractable obstacle in
turning a photo into a map, BUT . . .
it lets us view overlapping airphotos
stereoscopically (i.e. in three dimensions), and
better still . . .
its measurable. By measuring the parallax
difference of conjugate objects on photos, we
can determine their height.
9
Good aerial cameras have been around for some
time . . .

• Metrogon-lens cameras like this CA-8 were in
  active use till the mid-1980s.
• Their optics matched the Kelsh Plotters much of
  the U.S. was mapped with such distortions as
  the Metrogon lenses introduced, the Kelsh could
  remove in plotting.

10
They Made Great Photos
This 1939 photo of a portion of St. Croix State
Park typifies the 120,000 scale coverage of
rural Minnesota taken in the first generation of
ASCS photography. It was recently scanned,
together with photos from 1950, 1956, 1969, 1983
and 1991, to support historical study of the park
area.
11
Cameras and Films Are Better Now
This 80kb JPEG of a DNR color infrared airphoto
in the St. Croix Park area contains only .000005
of the total possible information content of the
original 9x9 film it was taken on. Thats
1/20,000th. Theres a lot of redundancy in modern
aerial photosthey contain more information than
youre likely to use.
12
Modern Aerial Cameras
The films are better, and so is everything else

• Resolving power greater than 100 lp/mm
• Forward motion compensation
• Distortion-free
• GPS-controlled shutter
• Gyro mount can be stabilized within a degree of
  vertical

13
If we want to turn an aerial photo into a map

• We must deal with --
• Camera orientation (tip, tilt, swing)
• Optical distortion, if significant
• Topographic displacement, if significant

If their effects can be reduced to within the
relevant map accuracy standards, were entitled
to call the result an ORTHOPHOTO
14
If tip, tilt and swing are the main problems . .
.
They can be dealt with by an expensive
photogrammetric projector with a table and lens
that can be tilted and turned to reproduce and
remove the effects of these camera
misorientations and bring the print to the
desired over-all scale. This is called PLANE
RECTIFICATION. The projector lens may also
compensate for camera lens distortions. But
topographic displacement remains.
15
For example --
Back in the 1950s some MN photo projects were
flown with cameras pointed obliquely fore and
aft. The negatives were then printed on a
rectifying projector to make sections come out
square at the desired scale. The frames became
trapezoids.
16
In the digital image world, rectification
is a term often loosely applied to a process
analogous to plane rectification, in which a
mathematical transformation is applied to rotate,
warp, stretch or rubber-sheet a digital image
to match a set of known ground control points.
As with plane rectification, this process (if the
transformation is properly chosen) can compensate
for systematic effects like orientation and
scale, but cant deal with topographic
displacement.
After transformation, the image must be
resampled to a regular array. The whole
business is perhaps better termed Geometric
Correction.
17
To handle displacement, we must talk about--
18
For purposes of illustration --
Heres an old Kelsh double-projection
stereoplotter. Most U.S. topographic maps were
created on this type of instrument.
Stereoplotters were invented to deal with
topographic displacement.
19
Heres how it works --

• When the projectors are properly adjusted in
• Interior orientation (same inside geometry as
  the cameras)
• Relative orientation (same tip, tilt, swing the
  cameras had)
• Absolute orientation (leveled with the mapping
  surface)

Then as long as the tracing table is kept in
contact with the surface of the stereomodel, the
pen orthographically traces the stereomodels
features onto the mapping surface. No more
topographic displacement!
20
Well, then --
If we can contrive to put a piece of film on the
tracing table, and keep it in constant contact
with the oriented projected stereomodel . . .
we can record a copy of the original photo bit by
bit, with all its topographic displacement
removed. The process is called DIFFERENTIAL
RECTIFICATION, because each bit of the photo gets
its own special treatment.
21
All sorts of really wild instruments . . .
were invented for doing this on the fly. Making
orthophotos this way brought a new world of
meaning to the term hand-eye coordination.
This is the USGSs original T-64 Orthophotoscope.
22
A French approach
. . . to the problem of orthophotoplotting is
seen in this Engins Matra model. Its kinship to
the Kelsh type of double-projection stereoplotter
is obvious.
23
Film table of Matra plotter
. . . shows the track in which the aperture moves
to expose the film.
Telescoping table legs keep the aperture in
contact with the stereomodel. The entire aperture
track steps across the width of the film to
record successive swaths of the orthophoto image.

24
Moving right along here . . .
-- because film was being continuously exposed!
As the aperture was mechanically cycled in X and
Y directions across the film bed, the operator
was on
his mettle to keep it continuously in contact
with the surface of the stereomodel by raising
and lowering the entire film table in the Z
direction!
25
Orthophoto Byproduct
And as successive rasters of an orthophoto were
scanned, the combined XYZ movements of the
aperture traced the topography of the
stereomodel. Once methods for recording these
movements were perfected, they could be turned
into contours or -- hey! -- Digital Elevation
Models! Minnesotas early DEMs showed
washboard-like traces of their derivation from
orthophoto scans.
26
This obviously couldnt go on . . .

• it wore out operators too fast. About this time
  computers came to the rescue --
• First, with off-line ortho scanning, in which
  the operator scanned the model at his own speed
  and played back the recorded XYZ movements to
  film the orthophoto.
• Then later with digital orthophotography, which
  turned everything on its head. Once the
  technology became available to move pixels around
  to their correct positions electronically rather
  than photographically, the DEM became, for most
  users, the driver of the process rather than its
  by-product. Orthophotos are now typically made
  by matching a photo with a pre-existing DEM in a
  computer. Of course the DEM still ultimately
  derives from some form of stereoplotting.

27
So now its all different --
And thanks to fast computers, all sorts of people
who never heard of double-projection
stereoplotters are busy creating orthophotos.
These days the necessary inputs (besides scanned
photos) are

• Ground control points
• Camera calibration parameters
• An adequate DEM for the area covered
• Photogrammetric software

28
Were skipping over important details
--e.g., project layout and control, which have a
great deal to do with final cost.
29
Expressing Film Resolving Power
30
So what?
Say you had a 140,000 negative at 100 lp/mm, and
wanted it all in your computer --
Youd need a scan aperture of 1/200 mm, or 5
microns. The pixels would measure 8 on the
ground . . .
And the file, for a color image, would be about
6.3 gigabytes.
31
How about orthophoto resolution?
How much can you afford? A 50 lp/mm negative
contains 1.5gb of potential information. Even an
800ppi scan (right, below) contains
only 1/10 of the total. Theres more data in
most airphotos than we can easily deal with.
32
Further Reading in Orthophotography
Demystifying Advancements in Digital
Orthophotography
http//spatialnews.geocomm.com/features/surdex1/