They do use aerial photos for the really zoomed-in bits, yeah. I don't think anything but the best military satellites could provide the kind of resolution you often see on Google Maps. Maybe not even then...
25 years ago, the KH-12 series of spy satellites reputedly had a resolution of around 2cm[1]. FAS also has a page that shows what different resolutions do with a newspaper headline[2]. 1 cm resolution can nearly read newspaper headlines from space (squint your eyes when you look at the 1 cm picture, and note that software can reconstruct that picture you see better than that method does).
Eh? The K-11/12 uses the same chassis as the Hubble.
Both of them have a 2.4M mirror, which means that under best conditions (minimal orbit) and no atmospheric distortion (which never happens) they'll have a max resolution of 16 or so centimeters (that's just basic optics for you), so not sure where did FAS got the 2cm number from. But i can tell you from 1st hand experience that even for military applications sub 50cm resolution is considered state of the art today.
US Spy satellites used to be always multi sensor platforms, the KH11/12 series was no different, and until the most recent ones they all had both digital and analog (film) components. 25 years ago the KH series digital sensor had a resolution of 400^2 pixel, maybe maybe in the mid 90's they managed to get a bird with 800^2 sensor.
The resolution is also dependent on the spectrum you use, SWIR is the most common in use for military satellites as it gives the user a larger operational envelope in being able to take imagery in low light conditions as well as penetrating light cloud cover and some non-adverse weather conditions.
However SWIR is just as much as affected by atmospheric distortions as visible light is, which means you tend to lose anywhere between 50% and 300% of your maximum resolution due to diffusion and refraction of the light within the atmosphere. (common back of the hand calculation is 2.5 times the maximum resolution of your optical system at MTF 50, which accidentally for the KH 2.4m mirror birds gives a resolution of about 40cm which slightly better than the best images released to the public from those platforms which was around the 50cm mark)
Thermal imaging tends to allow the sensor to capture much higher resolution image e.g. identifying individual people against a background (which is maybe where the 2cm resolution nonsense comes form) but those images are not graphically interpretable, they literally appear as white dots on a dark background.
SAR imagery probably gives the best resolution these days since it's virtually not effected by any atmospheric conditions and there are tons of ways to cheat with it to give you even sub centimeter resolution it just will take weeks if not months to take that image. SAR however will not allow you to read a headline or a license plate since it uses RADAR interpolation to construct a visible image.
Visible spectrum sensors are pretty much not used in military applications outside of modern aerial photography, and even then they'll always will have a SWIR or SAR fallback sensor. Simply because in a military application "color" isn't as important as being able to take pictures in low light to pitch dark and in all weather conditions.
That said the development life cycle for military equipment is much much longer than it's civilian counterparts, and in fact many earth observing civilian satellites tended to have better resolution than many if not all of their military (at least non-US ones) counterparts, many countries actually buy their satellite images from civilian companies EROS A and B (ImageSat) have about 80% foreign government utilization mostly European and APAC countries that lack the ability to launch their own birds, or have lacking coverage in general. So it would not surprise me if this bird has better capabilities than most current military sat's out there, they don't need to jump trough a billion hoops and standards, worry about compatibility with legacy and future platforms, or worry about congress cutting funding in the next mid term elections.
But damn 30cm per pixel in true color, thats better not only than most satellites out there, but than most high altitude atmospheric platforms as well.
Especially considering that no one besides the US has the capability to launch big birds these days, and even the US lost their ability to refuel them so not many 20,000 KG EOS's are going up there.
I'm really wondering what optics they use since I don't think they use a mirror which comes even close to the 2.4M the NRO was able to use. They must to have build a pretty insane adaptive optics setup to get that resolution going.
