[Beowulf] Re: blackbox on Mars?

Geoff Jacobs gdjacobs at gmail.com
Mon Oct 30 14:23:01 PST 2006


Jim Lux wrote:
> At 10:57 AM 10/30/2006, Robert G. Brown wrote:
>> On Mon, 30 Oct 2006, Geoff Jacobs wrote:
>>>>>> Mmm... Except... The high res images are from a plane because you
>>>>>> can't
>>>>>> really make out the fine details from a satellite through the earth's
>>>>>> atmosphere.
>>>>
>>>> It's true that most of the Google Earth images are aerial photos, but
>>>> I would imagine that one can get 10s of centimeter resolution from
>>>> orbit
>>>> on Earth (assuming that clouds aren't in the way).   An old
>>>> surveillance
>>>> satellite (Corona) was doing better than 2 meter resolution in the
>>>> 1970s.
>>>
>>> Absolutely. Assume modern recon sats use a primary mirror similar in
>>> size to the Hubble primary (both made by Perkin-Elmer). Calculating the
>>> Rayleigh limit for an Improved Crystal satellite such as that launched
>>> with USA 186 (Apogee 1050km, Perogee 264km) gives a max resolution of
>>> 7.9cm at 600nm and min resolution 31.5cm at 600nm. Space Imaging
>>> typically quotes max res. of 1m with IKONOS. DigitalGlobe says 61cm in
>>> B&W with their satellites.
>>>
>>> I would suspect that methods for dealing with atmospheric degradation is
>>> the secret sauce in the NROs architecture, especially WRT real-time
>>> applications of IMGINT.
> 
> With a nice braadband pipe back to the ground, and a modest rack of
> computers, amazing things are possible, including resolving things below
> the diffraction limit.  Look at the stuff people have been doing with
> imaging ISS and Shuttle from the ground with fairly small telescopes. 
> Indeed, the variable atmosphere can actually help, because it
> essentially gives you the ability to do multiple samples with
> statistically independent distortions, so you can "average" them to
> reduce the variance.  Something else to do with that Beowulf sitting in
> your garage..
I believe these techniques require multiple, short exposure samples.
Apparently, KH-12s have a capability for real time video. I expect the
frame rate from such would make multi-sampling a difficult noise
reduction method to apply.

> 
> I seem to recall that the ambitious amateur can get resolutions less
> than a meter with "available at retail stores" kind of optical equipment.
> 
> There is a ground based optical observing site on Mt. Haleakala to do
> just this sort of thing (or, at least, it's rumored to do so).
> 
> 
>> Awww, what you guys are all trying to tell me seems to be that I
>> shouldn't believe everything I see on 24.  So agent Jack Bower really
>> can't call back to CTU to track the driver of the grey mercedes from
>> where he abandons the car to where he disappears into the abandoned
>> military bomb shelter -- at night and independent of the LA weather and
>> smog.
> 
> IR imagery is one thing to think about.  Radar is another.  X band radar
> (9-10 GHz) easily gives you 3cm sorts of resolution, especially with SAR
> processing.
IR imagery does allow remote sensing at night, but it is affected by
weather. Also, IR typically has less resolution.

Radar allows all-weather remote sensing. Radar antennas are more easily
compacted for launch, too. However, you won't find Joe Taliban humping
the Hindu Kush with SAR (unless you can image his Kalishnakov).

>> You're really shaking my worldview here.  Next you're going to tell me
>> that Gil Grissom can't really prove that the sultry blonde did it from
>> the tiny splinter removed from the carpet at the feet of the victim that
>> could only have come from her imported chopsticks being hurled at high
>> velocity through the victim's brain...;-) Or that they can take the
>> blurred, crappy, low resolution picture from the surveillance videocam
>> in the parking deck, load it into their CSI Windows GUI and click on it
>> to prove that the perp was wearing argyle socks and had a mole on his
>> left butt-cheek by mysteriously increasing the available pixel
>> resolution by 2000% or so.
You forgot the part where they determine a particle of pollen came from
central Mongolia, which they then illustrate with a pretty little graph
using GIS data that, I guess, they just have on hand.

Then again, all this is about as realistic as having forensic
technicians do interviews of murder suspects.

>> I'll bet that MICROSOFT's CSI/beowulf software can do that and match
>> fingerprints too...:-)
But Macs are better at real life stuff... (wave hands vaguely)

> But of course.  You have the name wrong, though.  That's MS Vista Crime
> Scene Investigation/Clustering Edition, and the surveillance camera
> imagery will need to be processed through an appropriate Digital Rights
> Management system to make sure that you have properly licensed the right
> to superresolution processing, no?
> 
> Jim
> 
> 


-- 
Geoffrey D. Jacobs

Go to the Chinese Restaurant,
Order the Special



More information about the Beowulf mailing list