home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
vis-ftp.cs.umass.edu
/
vis-ftp.cs.umass.edu.tar
/
vis-ftp.cs.umass.edu
/
pub
/
Software
/
ASCENDER
/
model.tar
/
model-board-1
/
floppy
/
README.TXT
< prev
Wrap
Text File
|
1994-04-11
|
5KB
|
227 lines
README.TXT File
RADIUS Unclassified Data Set
Model Board 1 Data Release V1.1
November 11, 1992
The enclosed 8mm tape and diskette contain 28 image
files, revised "ground" control data files, and a number of
items requested by several of you. The Mac diskette
contains duplicates of all but the image files for those who
have Mac or PC systems for word processing. A printout of
the information files is also enclosed. The site diagram is
on four sheets to be taped together.
The 8mm tape was written by a Sun using the tar
command:
setenv TAPE /dev/rst0
tar c <list of file names>
The diskette is a 1.44 MByte Mac diskette and contains one
MacDraw file and a number of Mac/MS-Word and ASCII text
files, of which this is one. A list of the files and their
contents is contained in PACKING.LST.
The tape contains 28 image files. The original plan
was to take 40 images from model board 1. However, Oscar
and Bill requested that we change the model board and take
several images which would support a "change detection"
scenario, so we have reserved up to 12 images for this
purpose. These scenes will contain many more vehicles,
varying numbers of vehicles in finished goods inventory
storage areas, roof patches (to break up the eveness of the
roof gray-scale), and partial destruction of one building
from a fire. Work is underway now to produce these images.
A similar change detection scenario has been developed for
the second model board and will be incorporated into the
images taken.
Some of the ground control data has changed, as has the
format. The first column in the data table contains an
English description and reference to the MacDraw diagram.
On the MacDraw diagram of the site each building is numbered
and there is a pointer to the position of each ground
control point. This MacDraw file obsoletes the snap-shots
originally supplied. In addition, several of you requested
additional control points at ground level and additional
ground-roof measurements of the same vertical edge. Thus,
there are additional points in the V1.1 release not present
in the first (V1.0) release, as well as several deleted
points.
Several of you discovered that there were
inconsistencies in the original ground control data file.
Upon investigation, we discovered that the machine operator
who measured the model board had misinterpreted his
instructions and in many of cases measured the wrong points!
We re measured every point and believe the enclosed file
corrects these errors. However, if you find any point that
seems incorrect or unreasonable, please bring it to our
attention immediately. We apologize for the errors. You
are all aware of the difficulty recording and transcribing
400 random numbers accurately. (Note that it was not
possible to replace the model board on the machine in
exactly the same position, so there are small differences in
all of the measurements. Without moving the board we
randomly re measured about a dozen points and found the
values repeatable to the 2nd or 3rd decimal place.)
Several of you requested camera calibration data for
the camera setup used to take the images. This information
is contained in CAMERA.TXT.
Several of you commented on some unusual image
characteristics which result from use of a model board in
the Image Simulation Facility (ISF). First, it should be
noted that the "sun" in the ISF is not 93,000,000 miles
away. It is actually only about 20 feet (see CAMERA.TXT)
from the model board. Thus, the shadow angles will not be
entirely uniform from edge to edge. If your algorithm is
critically dependent on absolutely uniform shadow angles,
you should use the upcoming aerial photograph images of Ft.
Hood and Houston.
Second, the look and sun angle and azimuth contained in
the image parameter files are only accurate to +/- a few
degrees. These parameters were provided to give an
approximation of the image taking parameters, but if you
need exact values for these parameters you will probably
need to perform a camera resection using the ground control
data.
Dave McKeown has suggested that, in order to compare
results between laboratories, for a selected subset of
ground control points in each image there be a "standard"
control point to image pixel correspondence. While this
would certainly be useful, currently producing such a
standard correspondence is not in anyone's statement of
work, nor do we have funding to support its production. If
any of you plan to produce such correspondences and would be
willing to share the data, we would be happy to include it
on the next distribution.
Finally, again, please address any comments or
criticisms to me at:
Jim Green
3300 N. 23rd St.
Arlington, VA 22201
703-351-2801: office
703-351-7760: home