6
$\begingroup$

The premise here is that a government is already secretly using satellite imagery to record every surface of the Earth and automatically catalogues probable military assets and movements using image recognition software. Given the below constraints, could this technology already exist and be in use as of 2020?

The following constraints apply:

  • It must be able to record all of the Earth's surface frequently enough to reliably detect fleet and convoy movements.
  • It must be able to process imagery fast enough to keep up with what it is recording.
  • The satellites must be able to cache and transmit enough data to a central data center to keep up with what it is recording. (It can either send all the raw data to the center to be processed there or pre-process the data in spaceand just send its findings to Earth. Whichever is easier.)
  • It must record and save in high enough resolution for a human arial image analyst to be able to verify any findings.
  • It can only use currently existing technologies.

A best answer should also address the following:

  • Whether or not the project would be within the military budget of any modern nation or military coalition.
  • Whether or not there is an adequate timeframe between when the required technologies were invented and now to complete the project.
  • Whether such satellites could actually be kept secrete.
  • Any technological barriers that would prevent this project from existing yet.

Clarifications

My understanding is that surveillance satellites can already track these things in real time in places you are looking for them, and over time in a more global since, but I am asking about global real-time tracking.

$\endgroup$
11
  • 5
    $\begingroup$ Unless I don't understand the question, this has existed for decades... $\endgroup$
    – Muuski
    Oct 16, 2019 at 19:41
  • 1
    $\begingroup$ Why getting on a 3 letters governative agency's list by making bombs, when you can get on it by giving out the specs of their satellite surveillance program? $\endgroup$
    – L.Dutch
    Oct 16, 2019 at 19:41
  • $\begingroup$ @Muuski Not just using satellites to take pictures of things in specific places, but real time AI driven detection and tracking on a global scale. IE: the computer tells you when it sees someone starting construction a secrete missile silo on some remote island somewhere you previously were not worried about monitoring. $\endgroup$
    – Nosajimiki
    Oct 16, 2019 at 19:51
  • 1
    $\begingroup$ This has been done in real life for quite some time. What's happend in the last decade is that the price has come down enough so that commercial corporations can also afford to pay for it. (Hint: You don't have to record all the surface of the Earth all the time. All you need to do is record the places where the armies and navies shoud be, and react when they make preparations to move.) $\endgroup$
    – AlexP
    Oct 16, 2019 at 19:57
  • 1
    $\begingroup$ @Nosajimiki tracking the movements of all vehicles large enough to transport the missile would be less intensive than trying to monitor every square foot of Earth to see if anybody starts digging a hole. $\endgroup$
    – krb
    Oct 17, 2019 at 15:31

5 Answers 5

3
$\begingroup$

Short answer: If something like this does not exist now, it will soon.

The exact specifications of military-grade satellite imaging quality is kept vague on purpose however we can assume with a high degree of certainty that the US military has at least some satellites which are capable of sub-decimeter imaging resolution. If we extrapolate and make some safe assumptions, I'd guess that under optimal conditions, the modern military spy satellite can identify what people are doing from orbit and possibly even if they're armed with weapons. If there was say, a table set up with rifles laid out on it, someone reviewing the images would be able to identify that there are weapons present. The problem with these satellites is that they're somewhat rare and are usually used to look at specific targets.

In the commercial sector, global surveillance already exists. The company called "Planet" owns a fleet of microsatellites which it calls "Doves" and which are all in a sun-synchronous orbit. Each Dove is around 10 by 10 by 30 centimeters and consists mostly of a camera and solar panels. Right now, around 150 Doves and other Planet satellites are active and in orbit. Due to their sun-synchronous orbit, the swarm (or "flock" as it's called by Planet) can image the entire Earth once per day. This scan usually occurs sometime in the morning, between 10am and noon local time everywhere on Earth. The photos that this swarm produces are around 3 meters in spacial resolution with the newer generation reaching sub-meter resolutions (around 70 cm).

This means, right now, anyone with the means can buy imagery of the entire Earth that's less than 24 hours old.

Still a 24 hour refresh rate isn't quite fast enough to track military moments "live" while you could identify military bases, grab snapshots of convoys, and locate ships, you wouldn't be able to tell where they are going. Still, this is quite good despite the fact that Planet is a commercial company. If the military desired such a capability (and does not already have it), it would be quite easy for them to deploy such a system. It wouldn't even break the bank or even be that significant compared to the USA's military budget. If they wanted to, they could probably push such a satellite system without anyone noticing.

As for the questions about data storage and image analysis, those are all doable. While the amount of data that would need to be worked with and stored would be enormous, there is precedent for the Military to build supercomputers. It would be expensive but there's no technological or practical limitations that would prevent such a system.

$\endgroup$
2
  • $\begingroup$ Do you know about how much the Dove flock costed to build? If so, it should be pretty easy to math out a higher frame rate. $\endgroup$
    – Nosajimiki
    Oct 22, 2019 at 19:22
  • $\begingroup$ I could not find the exact cost of the flock, but based on the company's total venture capital, it would be reasonable to assume that a frame rate of just a few minutes could be achieved for about the same budget as the B-2 bomber program. Perhaps even in the seconds range when you look at economy of scale factors. $\endgroup$
    – Nosajimiki
    Nov 27, 2019 at 16:13
4
$\begingroup$

No way. Not enough satellites to do that.

I think the biggest barrier to this is the number of satellites required to do this. It also depends upon the kind of resolution you'd be looking for. If we're talking about Google Maps full-zoom level of detail of the entire planet, you'd need a number of satellites that can't possibly exist in the time since they've been invented.

Nowadays, satellites are used to take pictures of or track very specific locations in high detail. Doing the whole world in close to real-time is just not possible.

The amount of data couldn't be processed with the computers we have now. (at least not in real-time)

The total surface area of the Earth is 510,072,000 km2. Even if we had one pixel per square km, that would roughly equal a 22,584 x 22,584 square image, which might not seem too large in terms of processing at first.

An image like that, completely blank and compressed as a JPEG is still about 7.6 MB. As you increase the resolution and add detail, it would go up exponentially. This means the amount of data is going to be many orders of magnitude greater than a measly 7.6 MB.


UPDATE: It is true that the question did not explicitly ask for "the whole surface of the world" to be monitored and only asked for "the whole world's military movements".

Here's a hypothetical: If you have a satellite pointed at a military base, and two trucks full of weapons leave that base and move in different directions, how do you track them both with that same satellite without redirecting another satellite? What if this were to occur at nearly all bases at the same time? The answer is you can't.

The question did not ask to monitor individual bases, but their movements, which requires tracking on all of the locations where military movement might occur. Which could be anywhere the government decides them to go.

$\endgroup$
9
  • 5
    $\begingroup$ 8MB JPEG is a childplay. Astronomy deals with terrabyte images (they are RAW mostly, it would be tens of GB in JPEG). So, say 800MB for 100m per pixel is more than doable in real time even for a company of collage students. (Problem with sattelite coverage is still there, of cause!) $\endgroup$
    – ksbes
    Oct 17, 2019 at 7:29
  • $\begingroup$ Furthermore, GE full zoom is created based on aerial photography - not satellites. $\endgroup$
    – Erik
    Oct 17, 2019 at 7:53
  • 1
    $\begingroup$ Nobody does it this way. What the relevant organizations do is look at the relatively few points of interest. For example, the Russian fleet is normally at Murmansk, Vladivostok, Petropavlovsk, St. Petersburg and Sevastopol. You keep those ports under observation. For the fleet to move it would have to make steam, which would be blindingly obvious. Then you start tracking the fleet using your naval assets. $\endgroup$
    – AlexP
    Oct 17, 2019 at 10:00
  • 1
    $\begingroup$ While there may not be enough satellites now, such a project wouldn't be that difficult. Planet, a private company, owns and operates over 80 satellites which image the entire earth every day down to sub 1 meter accuracy. You can go to their website now and buy access. Sure the image is big but we have ludicrous computers. If a private company can image the entire planet every day with enough resolution to identify vehicles with only 80 satellites, the military could launch a couple hundred or thousand without breaking the bank and get a "global framerate" fast enough to track the military $\endgroup$
    – Dragongeek
    Oct 17, 2019 at 21:58
  • $\begingroup$ While the image would be big, It still wouldn't be "impossible" to work with. Assuming we image the entire planet with a 1-meter resolution, at 16-bit color, the image would be right around 1 petabyte big, uncompressed. That's big, but not that big. At "big data" companies, like Google, programmers are allocated multiple petabytes of storage space for free to store their personal projects. $\endgroup$
    – Dragongeek
    Oct 18, 2019 at 11:37
3
$\begingroup$

I think Overlord's answer is wrong. I think some big governments already have things in place like this.

First of all, we already have enough satellites to cover the entire planet. Look at google earth. Not enough for real time perhaps, but that is because Google is a company looking for profit. If they wanted to, they could assign satellites to this.

But, you do not need to cover the entire planet. After the start up phase, you will have a location on all military personnel, equipment and locations. And then it is a matter of tracking those. You can assign more satellites that cross the earth and scan it, to look for movement you have not logged yet. Aggregate this data with intelligence data, and adding any new movement to the network will be rather quickly.

I also want to point out that the images we see on google maps' satellite view are artificially made of a lesser resolution. The original images are so good that you could very well see a person in their home, through the reflection of a car mirror parked in the street. And these images can further be cleaned by software as well. The reason we do not see this is privacy laws and such.

We also have the infrastructure to process things like this. You just need to be smart about it. If you send all imagery in a data stream to a central point where it will get processed, it won't be possible. But such a system needs to be split up anyway. You will have cells of satellites working together, covering a certain area with their camera's overlapping each other's view. these satellites will have a small computational centre, in space or on ground, that will cover their images, tie it in a single feed AND perform calculations on that.

Again, you do not want to track every square meter of earth. You want to track specific locations and targets. That is what these cell's would do. Since these things are known beforehand, they simply need to track these. If stuff is not moving, there is probably no reason to notify anybody either. So the feed only needs to ring a bell if there is movement or activity happening. This simplifies the system greatly already.

Now all these cells will probably send their data to a bigger hub, which will send it to a bigger hub and so on. Somewhere all this data will be stored, but not in its entirety. Only the relevant bits.

Next to this you would have a separate network of moving satellites that are not in a stationary orbit. They will move across the globe and provide images this way. You can use software to let them look for certain things (and believe me, the stuff we can do with image recognition these days in commercial and hobby projects is amazing. What governments can do is always miles ahead of this.) and flag these. Another system will check these flags (with or without human input) and if desirable add them to the tracking network as a target to keep track off.

You can feed intelligence data to your second network and use that to look for certain things in certain places. eg a spy tells you about a secret base on some island in the pacific at these coordinates. Make network 2 search and locate it, flag it and then add it to network 1 to keep track off.

All of this would cost billions and you will need a large infrastructure. But this is certainly possible. The NSA alone would have the computational infrastructure to process this kind of data and they are only one of many intelligence and security agencies the USA have. If an entire country works on a project such as this; this is possible. If multiple countries (say the UN or an alliance of sorts) work on this, it becomes even more easy.

Heck, if Google or Amazon wanted to, they could do this probably.

If you are curious you should look up what people can do these days with image recognition software. I saw a tech talk from a CEO of a company that is using the windows UWP platform on vending machines in Africa, where they use small camera's to image recognise globally listed criminals. This is image recognition done on vending machine computers. The hardest part was not the image recognition either, it was sending the data over shit internet in the middle of nowhere. So they actively pushed all the processing and such to the machine. With smart algorithms this was possible.

And to get a feel of the difference in tech between what we civilians can use and what governments have, look up the difference between night goggles you or I can buy, and the stuff soldiers have access too. They have optics that can turn a pitch black night into a day light scene.

This applies to almost any technological industry.

$\endgroup$
6
  • $\begingroup$ "The original images are so good that you could very well see a person in their home, through the reflection of a car mirror parked in the street." That's a popular myth, but no, you couldn't, it's physically impossible, never mind our tech is far from perfection. $\endgroup$
    – Alice
    Oct 17, 2019 at 12:28
  • $\begingroup$ @Alice I have seen such an image during a tech talk though. So either that was tampered or it is not a myth. Now that I think back on it might have been an image taken from a drone/ plane and not a LEO satellite. $\endgroup$
    – Robin
    Oct 17, 2019 at 13:00
  • 1
    $\begingroup$ @Robin, A lot of good points on data handling, but I'm not sure about the cameras. Modern optics can zoom in far enough to do what you say, but only when focused on that one tiny spot. The larger of an area you try to capture, the lower the resolution. $\endgroup$
    – Nosajimiki
    Oct 17, 2019 at 14:00
  • $\begingroup$ But a single satellite does not need to carry only one camera. I also would guess that a system like this would not use images and look at them like we do. I assume they would use different wavelengths than what we can see. Also, the actual resolution and size of the first pictures of the moon were so good they had to print them out in large squares and lay them out on the floor and then still could go over them with magnifying glasses to look for the perfect landing spot. And that was half a century ago. $\endgroup$
    – Robin
    Oct 17, 2019 at 14:03
  • $\begingroup$ That is a valid argument, but higher resolution images like that also tend to take longer exposures and larger photo-cells. Any measure of image quality for this application should probably be measured in Area / (Framerate * Cost) @ a given Resolution unless there are other factors you can think to exploit. $\endgroup$
    – Nosajimiki
    Oct 17, 2019 at 14:16
2
$\begingroup$

In my opinion Overlord has it right. This is not possible (using only satellites).

Okay, let's tick them off:

It must be able to record all of the Earth's surface frequently enough to reliably detect fleet and convoy movements.

Repetition is an issue. Modern satellites can cover the same area within 10 to 14 days, if they take images continuously. Some platforms offer fast repetetions of about a day, but this is due to them being able to shift sensors to an area they passed over the day before. Also, optical satellites are limited, since they rely on sunshine and can't see through clouds. You can remedy this by using radar mostly and shooting up a lot more satellites.

It must be able to process imagery fast enough to keep up with what it is recording. The satellites must be able to cache and transmit enough data to a central data center to keep up with what it is recording. (It can either send all the raw data to the center to be processed there or pre-process the data in spaceand just send its findings to Earth. Whichever is easier.)

Not an issue actually. This is how it is done today: take images, save them, burst signal to ground, repeat - but it can't be done in realtime, since the satellites need a line of sight to a ground station which receives the data. You'll have a delay of at least several hours - and don't forget about the processing and analysis of the images.

It must record and save in high enough resolution for a human arial image analyst to be able to verify any findings.

Here's the main issue: Resolution vs. repetition. You either can get a high resolution, or a high repetition. Not both. The better the resolution, the smaller the area covered by the image, since sensors on a satellite only can be that large. But the smaller the ground area covered, the longer it takes to cover all area of interest, thus leading to a longer repetition cycle.

Still, you'd wonder what a trained eye can see on a blurry satellite image with 2x2 m resolution. Which is why many military installations have hangars in which aircrafts or tanks are parked. But this would mean that every image has to be analysed by a human - software still isn't up to the task (which is why captchas ask us to identify cars or streetsigns).

It can only use currently existing technologies.

Basically, if you plaster Earth's orbit with enough satellites, you could achieve this goal. But the amount of data would be tremendous, and the manpower needed to sift through all the images immens. Also, you couldn't keep it a real secret with all the rockets you'd have to launch.

What you could do:

  • Create a database of the really interesting areas (as mentioned in the comments) using satellites and update that database regularly (at least once a month)
  • launch some more satellites to increase your repetition (maybe every 4 days)
  • use drones and/or AWAC to increase your monitoring of the really interesting areas (as the US does today) - this would give you almost real time data
  • this still would leave you with a lot of data to sift through
$\endgroup$
2
  • $\begingroup$ "since the satellites need a line of sight to a ground station which receives the data" It is pretty common for satellites to communicate with each other and pass through data between ground station and a satellite opposite of it. While for most uses it has a limited bandwidth atm, some links provides GBps, so it is absolutely possible to make satellites networked with enough speed for real-time transfer. $\endgroup$
    – Alice
    Oct 17, 2019 at 12:34
  • $\begingroup$ Bots have been solving captchas for at least a decade now. Infact, they are much better than humans at it. ReCaptchas that are used today are actually biometric tests. They track how you navigate a user interface, how often you are filling out online forms by tracking you from website to website, etc. If your behavior falls within normal, it lets you in even if you don't solve the puzzle perfectly, but if you are crawling a bunch of sites filling out forms every 3 seconds, you will quickly be blocked across the entire reCaptcha network. $\endgroup$
    – Nosajimiki
    Oct 17, 2019 at 14:33
0
$\begingroup$

The technical answer is it would be possible if you devote an improbable amount of resources to it, as in tens of thousands of satellites in orbit and supercomputer farms with clusters of supercomputers wired together. Not to mention massive numbers of ground stations to handle the date streams coming in from space, and armies of analysts to interpret the data.

As the OPFOR, knowing this, I enlist my countrymen with an email. They report to training by taking the bus to the factory or office, and are escorted to the basement where a gym, range and classrooms are set up. They also have a social media account with us as well, to receive orders, pass on information and so on. They never wear a uniform in public, and after your three year enlistment you leave their "job" at the factory and go on to regular employment and remain in the reserves.

Want to invade a country? They receive a ticket via email, take the bus to the airport or train station and travel to your destination. We have already been preparing for years by smuggling in small weapons and equipment via diplomatic pouches, straw purchases in the local economy and so on. They and their fellow "vacationers" rent cars and travel to your destination, with a stop or two on the way to certain hotels or tourist destinations to pick up "items" they might need. The nation being invaded might even have Air B'n'b accommodations for lots of the "vacationers"...

So what is your satellite system going to tell you about my military or where or where they are operating?

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .