Re: [AP] Automated video-scanning for criminal mugshots

From: Eugene.Leitl@lrz.uni-muenchen.de
Date: Sun Feb 04 2001 - 22:29:27 PST


Strata Rose Chalup wrote:

> This kind of thing sounds reasonable, but I seriously doubt that the
> mug shot checker runs in realtime. It is extraordinarily unlikely

There are mug shot checkers which run in realtime, but that one
probably isn't (segmenting many faces in a crowd is not very trivial).

However, there is no reason why this thing shouldn't be able to
run in realtime if you throw a little more hardware at it (such as
a 19" rack full of Alphas, or a shelf full of Athlons), or
just wait until the next round of Moore's law strikes again.

> that the mug shot checker can in any substantial way reduce the
> risk of 50K people being killed at a Super Bowl. In fact, it is
> more likely to increase the risk, since someone might take it
> as a challenge. In any case, now that it is known that these

In fact I hope just that: as a challenge to wear silly masks in
public. The database only encodes facial geometry, body geometry
and dynamic patterns (gait, manneurisms) are absent, and are more
difficult to recognize anyway. We don't have to make it too easy,
right?

> things are being employed at similar events, if there were something
> planned, the planners would merely be more careful to use someone
> whose face wasn't known yet.
>
> Most likely, the mug shot checker is being used in the hopes of
> apprehending folks whose whereabouts are not known, in specific or

I see poor plastic surgeons pulling big overtime.

> in general. By getting at least one valid data tuple (so-and-so,
> was here, at this date), law enforcement has a starting point
> from which to track them down. If they get a match, they can then
> go and requisition camera feeds from surveillance cams in the parking
> garages nearby, the 7-11s, etc, in the hopes of tying the person to
> a particular vehicle, etc. They can check things they wouldn't have
> checked before.

All limitations of current technology. In theory, you could recognize
in realtime and track the person with a marking laser spot, so that
he's easier to shoot at. Of course, the next iteration of that technology
is too shoot the poor blighter immediately after he's recognized, and can
be shot at safely...
 
> So if someone challenges you saying, "Isn't the safety of FIFTY
> THOUSAND people worth a little privacy discomfort?", you might want
> to explain this line of reasoning to them. I'm not saying that
> the real results are or aren't worth it. I *am* saying that it's
> not at all a cut and dried case of "my privacy vs. 50K lives".
> If someone tries to put it that way, set phasers on "extreme
> rebuttal" and drain your powerpack at them.



This archive was generated by hypermail 2b29 : Fri Apr 27 2001 - 23:17:25 PDT