Hacker News Reader

Comment by pointyhat | Hacker News Reader

I worked for a couple of nefarious organisations over the years.

I doubt there is nothing much greater in resolution. There are so many technical challenges such as diffusion due to moisture in the atmosphere that limit things practically. The resolution is basically shite from space. Certain non-visible light bandwidth applications are better as they don't suffer from the same effects however they have to then deal with cosmic radiation (which is terrible apparently) and corresponding gaps in the magnetosphere which screw up the equipment.

It's a bit hit and miss really and costs lots of money for no real good.

If they had anything significantly better, they'd have decommissioned all the Lockheed U2's by now.

View on HN

The optical equipment I work with has improved so much since the 70s that I find it hard to believe spy sat technology has not.

I am not sure that the U2 is still used for surveillance. I think it is now used for research purposes. The much more modern Lockheed SR-71 Blackbird has been retired. Either the DoD is relying on sats or UAVs or the rumored next-gen spy plane is deployed [0]

Even if sats can take hi-res photos, there are still plenty of good reasons to use spy planes. Capturing visible light from space only works if the target area is not covered in cloud, spy sats are very hard to hide (it can be done in the short term but amateurs eventually find them) and they can be shot down (in a war with China, you bet the spy sat will be knocked out within hours of the first hostilities), spy planes can carry SIGINT gear as well.

That said, there are plenty of good reasons not to use spy planes and rely solely on sat intel. If you plane crashes it is a diplomatic and technological disaster. Unlike the stealth Blackhawk crash at Bin Laden's compound, spy planes don't carry a squad of seals equipped with C4 to destroy the valuable top secret technology.

[0] http://en.wikipedia.org/wiki/Aurora_%28aircraft%29

I think the limiting factor is the air, not the lens. That's why astronomers put telescopes in space (to avoid the air completely), or use a huge array of cameras and postprocess it.

I'd expect the big gains are in stitching together lots of signals, possibly over time, and extracting interesting features automatically. I'd think resolving stuff over time can only be done if you already know what you are looking for (to avoid massive dimensionality problems), or for static features. Still, that's an interesting subset.

Just like you can link a bunch of smaller telescopes on the ground to compensate for air distortion and get a significantly higher resolution, it should be possible to do the same with a bunch of satellites looking down.

But then you have to deal with parallax from nearby objects on the ground. That would limit its usefulness in hilly, mountainous or urban environments.

That's something computers are really, really good at.

The cameras don't have to be too widely separated. A little time and space helps a lot. The main problem is that this reduces your areal coverage as you're dedicating multiple sensors to a single sensing target.

Photostacking. Kalman filter. Differential.

U-2s are still in use, though they're in the process of phasing them out. They are still being used in a recon role (though no longer the same way as in the Cold War). They can still fly higher than Global Hawks, and there still aren't -that- many Global Hawks around.

It's easier just to jam the satellites directly instead of shooting them down. In the case of anything less than a full-on shooting war it leads to much less diplomatic trouble as well.

What optical equipment are you working with, if I may ask?

You can't use commercial focal planes in space. Rad hardened space hardware development moves slower than consumer.

In black and white, the GeoEye-1 satellite is capable of resolving objects on the ground as small as 16 inches. Average Joes like us don't see this imagery; it's sold to USG, while companies like Google are required to buy lower quality output.

That said, the "nefarious organizations" have satellites that can do amazing things...

Or so they'd like you to think.

UAVs get a lot more mileage than manned surveillance craft. Cheaper, smaller, fly lower, greatly reduced crew risk, and ability to swap out crew mid-flight.

Particularly interesting now are smaller drones -- from the size of typical hobbyist RC planes to insect-size micro-drones useful for surveillance in highly localized areas or inside buildings.

Eye-in-the-sky was really useful when we didn't have a good idea of where on Earth things were (the US made significant "corrections" to Soviet-era maps of central Asia in the 1950s and 60s). While they're useful to keep tabs on things, high-value surveillance tends to be more localized and focused.

Am I understanding this correctly that you're saying there is no equipment today that can take pictures at a better resolution than 1 pixel = 0.6 meter? Because even Google Earth does better than that, so I must be misunderstanding you.

Related: what am I misunderstanding when people talk about "satellites had a resolution of about 2 to 3 feet" and I think it's nonsense? How can a 'resolution' be '2 to 3 feet' - isn't resolution always something per something else - amount of detail you can represent onto another display surface? What part is implied here that I'm missing?

I believe most of the high res photography in Google Maps and Google Earth is taken using planes, not satellites.

When someone says a satellite or aerial image has a resolution of 2-3 feet, they mean each pixel in the image represents 2-3 feet on the ground.

A resolution of 2 to 3 feet means that features as small as 2 to 3 feet across can be distinguished, regardless of sensor technology. This is different from saying "2 to 3 feet per pixel" because the resolution of the lens and atmosphere may be worse than the resolution of the image sensor or film.

Consider a <$100 consumer digital camera with a hypothetical 12MP sensor (4000x3000 resolution). Suppose you point the camera at a 40' by 30' wall with a highly detailed pattern. Though the actual sensor resolution to wall resolution would be 0.01' per pixel, due to various optical distortions (and possibly the Bayer filter on the sensor), the effective resolution could be closer to 0.05'. Thus, any element of the pattern printed on the wall that was smaller than 0.05' would be unresolvable.

Google Earth uses both aerial and satellite images.

2-3 feet per pixel. Google earth uses that for imaging at very high scales. They use aerial photography for the rest. Per-pixel, it's definitely mainly aerial photography TBH.

what about fusing more than one lens? isn't the issue with resolution essentially one of variance, so by using a second, independent lens, one should be able to improve resolution by a factor of two or sqrt(two)? (not that it's so simple; registration must be a pain.)

I'm not sure about its space applications, but synthetic aperture radar has improved shockingly (I don't use the term lightly) over the past couple decades.

Yes it has but "it ain't that good". There's a lot of RF noise up there, so they still use air-based platforms such as Raytheon ASTOR/Sentinel. They're cheaper as well, even if you need a couple of Tornados to follow it.

I'm not saying any more otherwise I'll probably get bagged and tagged.

Reply on news.ycombinator.com