perm filename PROGRS.DOC[DMA,HE] blob
sn#686658 filedate 1982-11-08 generic text, type C, neo UTF8
COMMENT ⊗ VALID 00002 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00002 00002 ∂11-Aug-82 2315 TOB stereo
C00019 ENDMK
C⊗;
∂11-Aug-82 2315 TOB stereo
Boland called today to say that the funds for the
postdoc have been allocated. Syracuse should be calling
us soon.
I called Paglioroli, telling him that the proposal would
go Wed or Thurs by Federal Express. That seemed ok.
I want you to lead the stereo effort. I expect to take
part. Please set up stereo meetings and lead them. Also,
please make sure that reports are delivered as required.
It is important that we begin testing algorithms on data.
How do we stand?
****
∂23-Aug-82 1303 TOB
We don't include it in the survey. We include whatever we can get
freely without that restriction. Yes, leave it out of the tech report.
∂22-Aug-82 2133 HHB
Tom,
What is your understanding of our arrangement with RADC/DMA about
discussion of the CDC research results. The big report that our
critique of their 1981 work is based on begins with a note
fom Scarano saying `not to be freely distributed'. Inside it has printed
that we are (in fact?) in violation of `Internation Traffic in Arms
Regulations' in allowing me (or any other foreign national) to read
it. Do we include it in the Survey? It's in my thesis, do I leave
it out of the tech report my thesis will become? Can I cite it in
any other papers? Seems a whole lot of fuss for a decidedly inadequate
system.
Harlyn
****
∂03-Sep-82 1016 RAG Phone Message
John Boland, RADC called. If you get in before 12:30 Pac. Time,
call him back. He said you have his number. -- Anil.
Here 15 and 16 of September
****
∂03-Sep-82 1233 MAS
To: "@BOTH.DIS[DIS,TOB]" at SU-AI
On September 14 at 4 PM, rm 252
Steven Shafer from Carnegie Mellon will talk about:
"Straight Generalized Cylinders".
****
∂16-Sep-82 2250 TOB radc postdoc
Harlyn
John Boland asked that we make up a set of
tasks and schedule for the RADC postdoc. This
will start things off right. Think carefully about
it.
Also, he wants some sort of demo. The NASA sponsor
is paying for this part and may come along for the
meeting at the end. Some demo capabilities would be
extremely helpful, even if not the final/whole/total/
complete system. Please write out a draft quickly on
these so that we can talk on Monday and make a plan.
I want the stereo meetings to focus around meeting our
schedule for the stereo system. Yes, part of that is
making a schedule.
****
∂22-Oct-82 1026 MAS
∂22-Oct-82 0949 TOB
Concerning RADC equipment
It is understood that RADC intends to take back equipment bought under the
contract for stereo mapping. We have a comment on the situation. The
value of equipment depends primarily on software and on personnel.
When the contract is over, the equipment will have much less value for RADC
than for Stanford for several reasons. We will have software developed which
integrates the hardware into our system. The software will almost certainly
not be transferrable because it will be dependent on system and other
hardware configuration. We will have people who developed the
software and have used it extensively. Also, at the close of the ccntract,
two years will have passed, during which new equipment will be introduced.
RADC's best choice at that time will depend on availability of new products
and must balance the cost of hardware plus the cost of software with
functionality of the resulting system.
****
∂25-Oct-82 1012 TOB
Jake Scherer (Oct 25)
Boland is irate (with reason)
scheduled tasks
milestones
dates
proposed demo
****
∂26-Oct-82 0027 TOB
Jake Scherer (Oct 25)
Boland is irate (with reason)
scheduled tasks
milestones
dates
proposed demo
NASA AMES images
camera transform
digital terrain model
** image digitizing
** extended edge input
tablet
automated
** stereo station for ground truth
** simulating constructing and interpreting rules (TOB)
matching restricted by terrain model
can calculate horizontal errors from vertical errors
combine with knowledge of actual correspondence
** OTVs and horizontal surfaces (JMM)
calculate mappings between arrays for ground, and images
multiple resolution
** implement inference rules, OTVs, horizontals as ACRONYM rules (JMM)
** registration (Jeanric Meller)
epipolar transforms using Gennery's routines
using vertices by hand
demo
line drawings with correspondences
OTVs
horizontals
DHMs edges with same correspondences
stereo station
HHBs program
Tom-
I have the images of the NASA-Ames site, but they are
not digitized at this point. There is some considerable amount
of work to do to get the two images registered. I am terribly busy
trying to get Testbed documentation finished up; if you want to
send someone over to do the many trials that will be needed to
get an appropriate area of the film digitized, that's fine with
me. Mike Lowry knows about as much as anybody about this.
--Andy
-------
****
∂27-Oct-82 2335 TOB
RADC postdoc
Harlyn:
It looks very good. It also seems like a lot, but we can do it, I think.
We should keep track of what we said we would do, so that we can test
our prediction ability. It seems like a good idea to have some room to
overachieve. Otherwise, we will always be playing catchup. Or, another
way, if we always predict our limits, and if we err, we can only lose, not
win. Is there anything that we can not promise and still be ok, while working
full tilt toward it? If we demo it then, we will have a big plus. I am
thinking of OTVs. What do you think?
Tom
∂26-Oct-82 1757 HHB
Is this okay to send to Boland?
TIMETABLE FOR THE RADC POSTDOC MILESTONES AND DEMONSTRATION
-----------------------------------------------------------
Stanford Artificial Intelligence Laboratory
October 25, 1982
Our work on the postdoctoral contract started with building a test facility for
interactive experimentation on correspondence algorithms. Design exercises on
this led us to test implementations of selected elements of the mapping system,
particularly those concerned with the representational structures, the mappings
between them, and components of the evaluation function. Testing is being done
with both hand-synthesized data, and data obtained from an automated process.
Comparisons will be made of the results of three processes: the rule-based,
manual matching of data, both hand synthesized and automatically extracted data; a
computer-assisted human stereo matching process (utilizing the Stanford-JPL
stereo station); and an existing fully-automated Stanford stereo matching
process.
The stereo station will be used as well in determining ground truth of imagery
for which camera model information is not available. We are including image
registration in this phase of the research. Stanford has automated processes
for determining a camera model from stereo imagery. This will be extended here
to provide precision adjustment to the camera model using edge terminations and
vertices. Research will be ongoing in determining mappings between ground and
image arrays, at multiple levels of resolution. Rules for inference of
surface shape from monocular views of orthogonal trihedral vertices (OTV's) will
be implemented within ACRONYM.
A demonstration will be arranged where the various matching schemes will be
applied to a common data set (a building complex and/or the NASA-AMES sight).
The demonstration will be of:
a) the hand synthesis of extended edges,
b) rule-based manual matching of these edges,
c) rule-based manual matching of edges from an automated
process [Marimont 1982],
d) automated matching using Baker's system,
e) automated matching using Baker's system with Marimont's
extended edges, and
** f) computer-assisted stereo correspondence on the stereo station.
Milestones
----------
Implementation of digitizing facility
(primarily completed) Nov 1
hand Digitization of edges of Nasa-Ames stereo pair Nov 8
building complex "
** is this edge data?
Automated segmentation of Nasa-Ames Nov 15
building complex "
Stereo station analysis for ground truth
building complex Nov 22
Automated matching with Baker system
building complex Nov 22
Nasa-Ames Nov 29
Registration of imagery from terminations
and vertices: Nasa-Ames Nov 29
building complex "
Simulation of rule system for correspondence
Nasa-Ames Dec 6
building complex "
Simulation of rule system for correspondence
using Marimont edges
building complex Dec 10
Nasa-Ames Dec 10
Automated matching with Baker system,
using Marimont edges
building complex Dec 10
Nasa-Ames Dec 10
Implementation of OTV inference rules as
ACRONYM rules Dec 10
Demonstration of matching systems Dec 17
Hand synthesis Automated edge Automated stereo Stereo station Registration
of edge data extraction(Marimont) matching (Baker) analysis
Nov 1 | | | | |
Nov 8 X | | | |
Nov 15 rule-based X | | |
matching | | | |
Nov 22 | ____________/ \_______________ X X |
Nov 29 | / \ X OTV's in ACRONYM X
Dec 6 | / \ | |
Dec 10 X- -X X
Dec 17 Demonstration