Might as well attach infrared/visible light cameras to each satellite.
Then he could provide the Internet to all parts of the globe and have surveillance of all parts of the globe as well.
Save all video recordings for a week and use it to combat crime.
Hmmm... maybe give some thought to the resolution you'd need to do this. You'd probably need at least 1 pixel per square foot and a framerate of at least 1 fps. At that resolution, a standing person would be only half a dozen pixels, at most.
If we take just the USA, the land area is about 3.8 million mi^2, with a mi^2 being about 27.9 million ft^2. So, just storing surveillance on the US would involve capturing about 106 quadrillion pixels per second.
This poses a few interesting problems. First: the cameras. If 1/10th of the satellites were used to cover the USA, then you'd have to capture and process about 327 trillion pixels per second, per satellite. AFAIK, the largest commodity image sensors are about 50 MPix, requiring 6.53 million sensors per satellite. And that assumes they'd have no overlap, which they certainly would.
Next, let's look at what it'd take to process so much data. The NVENC engine, in Nvidia's Turing GPUs can compress about 740 MPix per second, in H.264. So, you'd need about 443 k Tesla T4 GPUs per satellite. Of course, these GPUs aren't hardened for satellite use, but let's leave that aside. Each of these burns 50 - 75 Watts, but let's take the low end of that range. So, you'd need at least 22 Megawatts of solar power generation, per satellite. If someone is familiar with the area-efficiency of the PV tech used in satellites, feel free to contribute an area estimate for the requisite panels. I think it's safe to say it'd be
large.
Now, a quick aside: we think of space debris as highly-destructive macro-particles - paint chips an bigger. However, there's a lot of microscopic dust up there, as well. This stuff has a sort of corrosive (or, to be more accurate, perhaps erosive) effect, leading to decreasing power generation efficiency and localized failures, over time. Also, the more
big satellites you have up there, the more targets for the larger space debris. Elon Musk says his StarLink satellites will be equipped with propulsion to help them dodge debris, but I think you're talking about some serious size and weight to equip these proposed camera platforms with such propulsion.
Finally, let's talk about the transmission and storage of all this data. Even 100:1 compression would mean about 100 GB/sec per satellite, assuming we start with 24 bits per pixel. I don't know anything about microwave communications, but I'll go out on a limb and say that's probably 2-3 orders of magnitude above their transmission capacity. And, on the ground, you've got to store that 3.18 petabytes per second, which would fill 318 10 TB HDDs
per second. That's 27 thousand enterprise disks (costing in the ballpark of $500, each), per day of video you want to store. Not to mention the cost of hosting and running them. If we assume the datacenter & server overhead brings it to about $1000 per disk, then storage for a week of video would be in the realm of $192 M.
So, this is
many orders of magnitude beyond the realm of feasibility. An interesting thought experiment, but absurd in the
extreme.
Now, if we restrict the surveillance just to densely-populated areas, you could probably lop about 3 orders of magnitude off those numbers, bringing it
closer to the realm of feasibility for a similarly-sized network of
dedicated surveillance satellites. Even for a DoD-level budget, it would still be quite a stretch. The reality is that spy satellites are just not cut out for such wide-area surveillance. You just can't watch everywhere, all at once.
One parting thought: what would you propose to do about overcast conditions or night time? If you do a pixel-level zoom on the raw image from a 50 MPix sensor, it's quite noisy, even in good light. If you take long exposures to deal with nighttime, then any moving objects are going to be blurry smudges. And thermal imaging won't save you, because thermal image sensors are way lower-resolution (and much more expensive) than optical.
BTW, feel free to check my math. I did this rather quickly, but I'm quite sure any errors I might've made didn't have much effect on the conclusion. I was fairly conservative on compression, but remember that the quality has to be good enough for the fully zoomed-in images to be useful.
Facebook
Twitter
Instagram