Sunday, September 7, 2008

Spy Satellites Chasing Shadows: NASA's Contribution to the War on Terror

995_sukhoi Are you having trouble with funding? Is your research unable to attract major media attention? Just add Terror (TM)! That's what Dr Stoica of NASA's Jet Propulsion Laboratory did, and it could work for you too.

Dr Stoica's research is based on gait analysis - the idea that everyone has a distinctive walking pattern, and no matter how many fake beards or dark glasses you put on you can't disguise your stride unless you're wearing so many your knees break.

So far so good. Things only go off the rails with his proposal to use this software to scan shadows to find terrorist suspects. From space. Seriously. Satellites have a hard time recognising individuals from the tops of their heads (the lack of Superman in our world means people rarely have reason to look straight up), which is why Stoica has developed software which can reconstruct their profile from their shadow and the time of day.

That sort of thing might work in CSI: Hawaii Beach (or whatever they're up to now), but here in the real world there are a few problems with the pitch:

1. Satellites simply do not have the required resolution. The highest commercial satellite resolution currently available is half a meter, meaning you can only make out details about that big. If your terror suspect can still be recognized at that scale you could have saved time by checking the International Sumo Federation database. The military often has better gear, of course, but in this case it's not a question of money - it's a question of what's physically possible.

2. Computer reconstruction of images might seem standard for the average movie-goer, but in the real world you can't just say "enhance" and capture a fingerprint from a week old pizza box. This idea calls for processor intensive, time sensitive reconstruction of hundreds of frames of footage for millions of people, all on the off chance of getting something clear enough to identify someone. There are a huge range of amazing data-mining applications of modern surveillance, but this isn't one of them. This is more a "with these new steam-engines, we could build a machine to care for infants" idea, i.e. a really bad one.

3. Walking patterns are distinctive, but they aren't "six billion different types" distinctive - especially not with reduced satellite resolution and when they've all been reprocessed by the same reconstruction algorithms (which will have their own bias).

4. Little thing, tiny point, but Dr Stoica does point out that this system will only catch those terrorists whose walking pattern is already on file. Unfortunately, the world's intelligence agencies don't actually have a catch-and-release policy for major terrorists - getting them, measuring them, then setting them loose to see which clever scientist can find them again - so we can expect this to be what we call a "major issue."

5. As well as the "it's good for Terror!" warning sign, there's also the "at the earliest stages of development" alarm bell in the announcements. Translation "You will see nothing useful out of this for ten years, if ever, and the only reason we're even mentioning it is we need the money/attention".

You'd be better off finding which soft drink terrorists like best, then tasking satellites to track the delivery trucks. At least you can SEE those from space.

Posted by Luke McKinney.

Original here

No comments: