Thomas Billings wrote in message ...
In article ,
(Bill Bogen) wrote:

It was once proposed to do an automated search of images from the
Clementine probe for evidence of lunar lava tubes. The argument (by
Taylor and Gibbs) was that the usually-cited candidates for lava tubes
are the huge, apparently partly collapsed rilles visible from Earth or
in Apollo photos and that any tubes at these sites may be too deep to
use easily. The thought was that smaller tubes could be just as
useful (for radiation and meteoroid protection), more numerous,
possibly located near more interesting sites, and easier to access.
These smaller tubes might be discernable in some of the 1.9 million
Clementine images.

Did anything ever come of this?

Our research team at Oregon L-5 attempted to start preparing for such a
search, back in 1993-4 using some software from a Caltech/JPL project in
pattern recognition. The originating software team had used it to search
out small volcanic features in the Magellan Radar data, and we thought
of using it on Clementine data. We knew it might be marginal, because of
the resolution of the Clementine sensors on the lunar surface, but
figured it was worth a try.

Unfortunately, the software (unnamed, to protect the guilty) turned out
to be such an unusable lash-up of previous academic projects that we
never found anyone outside of that particular Caltech/JPL team who had
gotten it to work, either. We once met someone at a conference who also
tried it, and he was awed that we'd once gotten as far as getting a user
interface screen! We banged away for about 5 years, on and off, as
volunteer teams must.

Any idea how many person-hours were consumed?


The head of that software project quit Caltech and joined Microsoft in
the middle of our efforts. 'Nuff said!

If not, did anyone ever begin a
_manual_ search of the 620,000 high-resolution visible-light
Clementine images?

We looked at this, and looked at our local support group, and quailed!

Why? Isn't the human eye&brain a wonderful pattern recognition
device? Maybe I'm being naive here but I'd set up the project this
way: 10 people each sitting in front of a PC for about 2 hours a day.
Display an image. The person decides whether it is a possible
lavatube and flags it with a keystroke. Since the vast bulk of
pictures will be rejects, I'd expect an average rate of about 1
sec/image. We'd be done in 9 days. Let's triple that and let each
image be seen by 3 people; we'd rank each image by consensus. Let's
pay each person $10/hr: labor cost = $5,167. Even adding costs for
software to present the images and record flags, project management,
etc, this still seems pretty cheap.

Oregon L-5 has some experience in NASA grants, doesn't it? Would this
be an unreasonable grant proposal?