2004
1etric
t did
| film
come
ssed:
what
now
tor of
bject
extra
e-lap
Ffects
and
is à
uting
mage
ation
, and
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
Figure 1 summarizes the advantages of digital sensing over film
images in photogrammetry. The advantages are grouped into 5
categories. Of course they are all interdependent. Reduced manual
labor affects the economy, yet it also represents an advantage in the
workflow. An increase in accuracy, while of value all by itself, also
affects the economy.
A very important factor is the ability to enjoy a no-cost increase of
image overlaps. This is denoted as "increased redundancy". This
factor is the least-often discussed when reviewing the transition to
digital cameras, yet it is considered by us to be the most
fundamental. Photogrammetry has been relying on two rays to
intersect a 3-dimensional point, basically without any redundancy. If
a ray from one image is in error, then the point position cannot be
measured at all, a deficiency of stereo-photogrammetry since its
inception. Insufficient redundancy is the major reason that
automatically created DEMs using stereo matching of scanned film
images have significant errors and miss data, and occlusions due to a
lack of redundancy are the major limitation of orthophoto products.
The transition from minimizing the number of film photos to
maximizing the robustness of automation by a free definition of the
number of images can be a major shift in the photogrammetric value
system, and can be called a , paradigm shift".
4. AUTOMATION OPPORTUNITIES
4.1 Multi-Ray Matching
Increasing the redundancy leads to a change from 2-image stereo
operations to multi-image, multi-ray matching. Aerial triangulation
tie points get collected from 10 images rather than 3 or 6 in model
and strip overlaps, leading to as much tighter fit and more robust
solution. DEM matching is with 10 images versus 2, and producing
an over-determined solution where previously there was no over-
determination. Orthophoto texture derives also from 10 versus only
2 inputs, and the need to carefully delineate the seam between
adjacent photo sources in one single orthophoto-mosaic goes away.
All this is simply from a transition to the 80/60 overlaps versus the
traditional 60/20. The robustness of matching increases. Dissimi-
larities between adjacent images get smaller and matching errors are
less likely. As a result, AT and DEM-creation become fully
automated, and the orthophoto will be “true” with no extra effort.
Can we hope to soon achieve softcopy direct geopositioning and a
softcopy terrain scanning function? Figure 2 presents an automati-
cally created DEM from 4 images. No manual work was applied, yet
there are no spikes nor holes visible in this DEM.
4.2 Use of Multiple Color Bands
Color had been a low priority issue in traditional photogrammetry.
Only the advent of the orthophoto has made color an important
factor, but all geometric processing is color- neutral.
Now we have a no-cost, free ability to increase the use of color
bands, starting out with at least red-green-blue-infrared. This will
improve the ability of automatically interpreting the contents of the
image. Image classification promises to become a routine operation
in the photogrammetric work flow to support all its operations, be it
for AT, DEMs or vector collection.
Figure 2: DEM created automatically by multi-ray matching. Each terrain
point is in at least 4 images. Above: 3 million triangles & image texture of an
area in Colorado Springs. Below: Detail with business building, highway and
parking lot with cars. Note the curved sidewalk along the parking lot.
DIGITAL CAMERA ECONOMY, W/O PHOTOGRAMMETRY
Item US$
Camera annual depreciation 80,000
Camera annual maintenance 40.000,
Camera miscellaneous (e.g. insurance) 10,000
Level-2 and 3 post processing depreciation 4.000
Archiving tapes, 60,000 digital images 2,000
Archiving & cataloging system, depreciation 30,000
Flight operations, extra for 60% sidelap flying 50,000
SUM 216,000
Per each of 60,000 digital images 4
Per each of 20,000 (film-equivalent) images 11
FILM CAMERA ECONOMY WITHOUT PHOTOGRAMMETRY
Item US$
Camera annual depreciation TBD
Camera annual maintenance 5,000
Camera miscellaneous (calibration) 5,000
Film purchases, 20,000 images annually 80,000
Photo processing, 20,000 images annually 60,000
Scanning, 20,000 images annually 200,000
Film archive operating costs, annually 1,500
Digital archive operating costs, annually 1,500
SUM 353,000
Per each of 20,000 film images 18
Figure 3: Direct costs creating single film (below) & single digital image
(above), based on annually 20,000 film and an equivalent 20,000 digital
images (when in fact 60,000 digital images get produced to cover the same
area). Digital coverage not only comes for less money than film, it also
comes at much higher overlap/redundancy. All costs assumed to be typical
for a mix of European and US operations, over all scales. Note that of course
costs may vary greatly with scale, area and organization. Basic flying costs
excluded.
761
