Software stereo-restitution

[CJMCE@xxxxxxxxxxxxxxxx]

To: photo-3d@xxxxxxxxxxxxxxx
In Particular: branger@xxxxxxxxxxxxxxxxxx (Hubert Branger@xxxxx)
From: cjmce@xxxxxxx (Cliff Mugnier @ University of New Orleans)

Subject:  Software stereo-restitution

The answer to the many questions posed is yes and no.  Yes, software exists
for the application, but only some of it is free under certain conditions.
The general topic I can address is photogrammetry.  I know what a Traster T10
is. Since you are so concerned about the cost of stereorestitution; you must
be using a commercial company.  If it is impractical for you to purchase your
own Trasta T10; what kind of computer do you have?  Are you looking for an
analysis tool to run on an IBM-compatible PC?  Are you looking for something
to run on a Mac?  Are you looking for something to run on SGI or Sun ?

If you really mean what you say, and all you want is XYZ coordinates of the
object you observe from the left and right images; there is hope for you in
the "PC" world and some UNIX environments.  Let's assume you can obtain the
pixel coordinate of your point (x,y) on the left image and the pixel
coordinate of your conjugate point (x,y) on the right image.  If you can do
that, then you have some sort of imaging software already running that will
allow you view the images and capture coordinates.  The next step, then, is
to perform photo-triangulation to obtain the XYZ coordinates in object space
from the conjugate xy coordinate pairs in image space.  These pairs of points
must be at a minimum of six, preferably nine (or more) per stereopair to
provide sufficient geometry for a Least Squares solution including analytical
resection and error propagation of the geometric dilution of precision (GDOP).

If the above "ifs" are satisfied, then what you might need is a commercial
photo-triangulation software package called PC Giant.  I wrote that package
and I sell it commercially.  However, if you teach photogrammetry at a
university, I offer a free "demonstration" version with documentation to
university professors.  If you write to me at my commercial address listed
below and use your University Letterhead requesting a free demo copy for
evaluation; I will send a free copy with the manual (164 pages) and a
copyright release allowing you to make unlimited copies of the manual and
demo disk for educational use.  The commercial price is $1,995.00, which is
inexpensive for such software with additional discounts for educational
institutions.  PC Giant is written for use by photogrammetrists.  It is not
easy for laypersons to use; it is almost incomprehensible...  If you know how
to calibrate a CCD camera for radial lens distortion, principal point of
auto-collimation, offset of the principal point, equivalent and calibrated
focal lengths, etc.; it sounds like you could use this demo copy.  (The
educational demo copy includes software for analytical camera calibration by
plumbline as well as trilateration network adjustment for use with
terrestrial, close-range and Forensic photogrammetric applications.)

This offer is open to anyone else on the FACULTY of a University that would
like a free demo copy with reproduction copyright releases.  The free offer
applies whether a production copy(s) is/are purchased or not (no strings). 
The same package is suitable for Aerotriangulation and comes with Country
versions.  I currently have demos available for USA, Canada, Spain, Ecuador,
Saudi Arabia & Australia.  Additional country-specific versions will become
available as production versions are purchased.  The manual is currently
available only in English with some minor documentation written in Spanish. 
The Spain and Ecuador versions do have the software written to use run-time
error messages in Spanish.  In spite of my last name, I cannot write in
French.  I do require letterhead requests; email requests are NOT honored
because my email address is not commercial.

In reference to the specification that precision required is 1 mm for an
object 60 cm by 60 cm, that is trivial if you provide the requisite geometry
and image space precision.  I work with Kodak DCS 460 cameras that use a CCD
array of 2k by 3k. I use three of them with three 60mm, 28mm & 18mm lenses
each.  (I'll bet there's a 3D Buff or two out there that wouldn't mind being
paid to play with my toy chest...)  Anyway, I easily get 1 mm precision &
accuracy in 15 METER parallelepipeds.  I do that with highly convergent
shots, and then oftentimes multiple triplets at that for complete 360 degree
coverage.  If you are doing "true" stereo, the precision and accuracy will
decrease to some variable as a function of your stereo base with respect to
your distance to the object. (Base/Height ratio) "Hyper stereo" to those
readers who prefer English.  Since you say you obtained your imagery TWO
years ago, you apparently do not have control over your geometry.  What
results you will get with PC Giant is not dependent on PC Giant; it's
dependant on what you did two years ago.

In reference to your request for an orthoimage, in Photogrammetry we term
that an Orthophoto Map.  For that, it's not free.  The least expensive but
fully robust package for that task (as well as Professor Crowell's comment
about the identification of conjugate points) is R-WEL's Digital Mapping
System (DMS).  R-WEL stands for Roy Welch @ the University of Georgia in
Athens, and he has educational pricing for his software. ($7,500 for a
package of 4 seats, and his software, just by coincidence, is fully
compatible with PC Giant.)  You can even buy PC Giant from R-WEL that works
with an R-WEL dongle.  The DMS package will allow stereo on a PC in anaglyph.
Professor Welch does not offer free demos of DMS.  Other software exists also.

Clifford J. Mugnier  
cjmce@xxxxxxx
Department of Civil and Environmental Engineering
University of New Orleans
New Orleans,  LA  70148
Voice: (504) 286-7095

Commercial:
GPA Associates
UNO Box 1200
New Orleans, LA  70148-1200
Voice: (504) 286-1200




------------------------------