PDA

View Full Version : SUMMARY: Video Analysis In *COLOR* papers/references?



Daniel C. Richardson
12-07-1994, 05:27 PM
(Sorry this is so late in coming, but...)

Howdy all again.

After a lot of looking, searching, and wondering, here finally are the
results of my quest for color video analysis systems:

(There ain't much out there)

The only paper that we found was on from SPIE Vol. 1356 IMAGE-BASED MOTION
MEASUREMENT (1990). It describes a system which uses color camera(s) and
up to 6 color markers. A call to the company revealed that it's '486
based, and hasn't been updated sinc1992. The engineer said it's "reliable
for 3 markers, not definate for the other 3". It's a decent "paper"
describing the companies VideoMex-X system, which is almost identical to
the experiment we rigged up a few weeks ago. Anyway, they're:
Columbus Instruments Corp.
950 North Hague Avenue
Columbus. Ohio 43204 USA
===

Other than that, I received many replies, but mainly from people offering
to help with image processing. Some saying that color wasn't quite up to
snuff for good imaging.

I've been so long in posting a reply listing because I wanted to wait for
the results from my experiment. We set up a PEAK system and COLOR system
(Mac Quadra, Video Vision Studio capture board), grabbed a set of data
where markers crossed over each other at various points (fully covered, 1/2
covered, barely touching, and 20 mm apart). The results were mixed, since
there was some question about our PEAK analysis, but I was able to
selectively eliminate all but the one color I was looking for (the markers
were pink spheres crossing behind green spheres). It was manual, tedius,
and not recommended for anyone wanting to keep their sanity, but it did
work. I think a fully automated system of data capture and analysis is
possible using relatively inexpensive hardware (well, under $50 grand...).

If anyone wants a copy of the results, e-mail me.

Oh, and if anyone is interesting in funding a project for the next 4 years,
I'm soliciting offers... ;-)

-Dan
=== My original posting ===

Howdy, all.

I've been searching the literature for ANY references to research that has
been done with color video cameras for motion analysis. So far, I've been
able to find plenty on B&W/Greyscale systems, and plenty on color re:
surgery (blood flows, etc..), but nothing on color.

I'd also be interested in papers on methods that DON'T use thresholding as
a main image processing as a main technique.

The reason I'm searching is that I'd like to test the hypothesis of whether
color is more {accurate, easily automated, reliable, etc...} than Greyscale
for motion analysis.

Thank you,
-Dan

=== Replies follow ===
Hi,

I saw your mail, but I wasn't sure of exactly what you are looking for. If you
are looking for general ideas about how one would do something like you've
described using digital imaging, I may be able to offer some pointers. If,
however, you are looking for pointers on how to use Premiere and Photoshop,
or how to program the mac, I'm not the right guy (I'm just learning to program
the Mac myself).

Anyhow, let me know if I can offer any imaging suggestions. I used to do some
of that sort of thing.

Frank
=====
Hi Daniel,
I'm not quite sure what youre saying, but I'll spout for a bit and see
if I'm on the right track. What I think youre describing is color keying,
tho I may be wrong. Color keying (or Chroma Keying) is a technique in which
a particular color is used to represent a mask. A common use is TV weather,
they have the forecaster standing infront of a large blue wall. Then you
take this signal and send it to a special effects unit, you also send
another signal that you want as the backround, in this case it would be the
satellite map, or a sun with a smiley face, etc. Then on the chroma keyer
you basically set it so it takes blue (there are usually two knobs,
controls, one for hue, another for Q, how closely the color has to be to
the target for it to be keyed). Then everything that is blue is replaced
with backround channel of video that you are sending. I believe Premiere
has a Chroma Keying Plug in, if not Gryphon might sell one. ALso CoSA After
Effect shas one, but it's sorta pricey.
Hope this helps,
Chris P
cdpro@student.umass.edu
=====
Hi Dan: Once upon a time ELITE attempted a product that did motion capture
based on marker color. They do not make it any more. You might want to
contact them to discuss why they gave it up.

=====
Dear Daniel C. Richardson
All the systems I know tend to use a threshold system(irrespective of the
colour B/W system) so if there are other systems I'm talking out of my
league and ignore the rest!
If you are concerned with a specific target then I suggest selecting a
frequency of light that specifically identifies that target. eg red light.
( I'm not sure how you could "track blood" uless you put some type of
marker which has a consistent boundary definition) Anyway, standard systems
can be used with the appropriate (red) filter to increase the contrast
between the background and the reflection off the target.
Cheers

Garry T Allison
School of Physiotherapy, Curtin University. Selby Street Perth
Western Australia. Phone(09) 351 3660. "G.Allison@info.curtin.edu.au"
1. Unauthorised individuals have access to this terminal please request
verification of author if necessary 2.These are my opinions,they are not
necessarily shared by others :-)
=====
Dear Dan,

Sorry, I can't help you with colour video analysis. I'm just starting out
trying to use B&W video cameras for motion analysis of hands. As you said
that you'd found a few references on this, I thought I'd be really cheeky
and ask you to point me in the right direction.

Thanks for your help!

Nicky Shaw-Hamilton
Bioengineering unit, University of Strathclyde, Glasgow.
n.shaw.hamilton@uk.ac.strath
=====
You have requested information on color motion analysis on Biomech-L group.
Prof John Ayers of Norwestern Univ. had written an article in Advanced
Imaging magazine , Nov. 1990 issue. The title is "Color, Motion Analysis
and Biological Imaging'. He made color extension to NIH-Image program which
works on Macintosh computer and both ColorImage and NIH-Image programs
are available free from zippy.nimh.nih.gov in /pub directory. ColorImage
may be in /conotrib subdirectory. You can use free software Fetch to
download the programs. Ayers may be reached at (617)581-7370, or he can
be reached through MacSciTech.
lichunwu@pioneer.arc.nasa.gov
=====
Hi Dan... I'm responding with respect to your request, but i'm afraid i
don't have any references for you. I, too, have been searching for
colour video imaging references to no avail. My area of interest is not
motion analysis, but i'd be very interested in any responses you
receive... could you post them to the list??

Thanks very much, and good luck.

Sherri Sparling
Mechanical Engineering
Queen's University, Kingston
Canada
=====
>From a video engineering point of view, the use of b/w cameras is superior
because of the colour encoding systems used (NTSC, PAL or SECAM). Since
the cameras and vcrs in use are those of the broadcast systems, the
compromises that were made when the broadcasting of colour began apply:
The signal was to be compatable with existing b/w receivers and must fit
into existing broadcast frequency allocations ie no extra bandwidth. These
aims were achieved by generating a b/w signal (Luminance or "Y" signal.)
from the RBG camera tubes. Colour difference signals were then produced
(R-Y and B-Y), reduced in bandwidth to about 1Mhz and phase and amplitude
encoded onto a sub carrier, within the 5Mhz. video bandwidth and chosen so
that the modulation had minimum effect on the high spatial frequency
detail. This signal, as NTSC and PAL is still used, and will be until at
least 2006 (the official death point of NTSC.)
The VCRs in general use are designed for this system and in fact all
degrade the signal further by extracting the colour sub-carrier,
down-converting it, reducing bandwidth to 0.5Mhz before recording it on
tape. The noise level in the colour section also increases.
By using the b/w, luminance signal only, the bandwidth is the best possible
eg 3 to 4 Mhz.or 3 to 400 "lines". for comparison, a studio camera has a
bandwidth RG and B, of 6Mhz. or 700"lines"
New digital technology will reduce the compromises but the machinery is
still very expensive.
The reason that the compromise works at all is that the human eye is not as
sensitive to colour changes as to luminance ie spatial bandwidth is lower
for colour than luminance. You may find a Television text book useful on
this subject.

Thresholding is necessary in auto-digitising systems to allow the location
of markers. You may like to look at some of the early work done on cine
film and manually digitised. Note that the PEAK system allows full grey
scale when manual digitising. It is awfully tedious but allows choice of
points that may not have markers.

>
>The reason I'm searching is that I'd like to test the hypothesis of whether
>color is more {accurate, easily automated, reliable, etc...} than Greyscale
>for motion analysis.
>

I would suggest not because of the perceptive and engineering limits of colour.

=====

-------------------------------------- KLR650 "The Dirt Ninja" --
|Daniel C. Richardson |Penn State University Graduate Studies |
|122 East Irvin Avenue |Industrial Engineering - Human Factors |
|State College, PA 16801 |"Education should not bore. Work should |
|(814) 234-6640 | not stress. Life should not pain." |
-- dcr126@psu.edu -----------------------------------------------