The best kittens, technology, and video games blog in the world.

Saturday, June 12, 2021

Is stealth even possible?

Previously I solved the Mechs vs Tanks question. Now let's move on to the next one - is stealth even possible, specifically on military aircraft?

Aircraft is usually detected at long distances by radar. Stealth aircraft has been a thing since 1980s - it uses various tricks to make it harder to detect over long distance by typically used kinds of radar, and that's effective enough.

There's been many attempts to use different kinds of radar systems to detect stealth aircraft. They're probably at least somewhat effective, but this information is not really available to anyone. Not even in the sense of being some military secret - how well Chinese detection technology can deal with US aircraft is unknown to both China and US, and likely won't be until China invades Taiwan and needs to test in it practice.

Phone cameras

However, why limit ourselves to radar? Let's do some back of an envelope calculations.

Military aircraft is about 10-15m long and wide, and moves through open space at very high speed. That's not like any natural object. Let's approximate that to 10mx10m square.

Modern consumer phones have cameras of about 100MP, or 10000x10000. If we imagine that it covers 90 FOV, then if it points at a 10km by 10km area, 10km up, each pixel covers about 1m x 1m - so aircraft will cover about 100 pixels and will be trivially detectable by even dumbest Python image recognition script, at least in daytime.

Let's assume we want to defend a big country like Iran from people fed up with its terrorism. Iran has area of 1.65 million km². As our phones each look at area 100km², spreading them so every point in the sky is seen by 10 phones at the same time - for huge redundancy, it would take just 165k phones, or at $500 each just $82m, less than one jet.

If we require lower coverage multiples, phones get better or cheaper, or we can detect enemy aircraft from less than 100 pixels, that drastically reduces the cost.

One might think that horizon would be a problem, but phones in this example are looking up, and horizon is 5km away at 2m tall pole, and 10km away at 8m tall pole.

Various terrain features like mountains are a problem for radar system - as radars are big and expensive and therefore very few. But we can cheaply install so many damn phones this problem is absolutely trivial.

Image Processing

Someone not familiar with modern image processing might imagine that detecting aircraft from a picture would be hard. Nowadays this is honestly a ridiculous idea. You can create fake tanks to fool the AI, but there are zero big fast moving objects in the air other than aircraft, and once you know aircraft is there, it really doesn't take much to figure out if it's a friendly, neutral, or enemy kind.

I can also think of no reason why similar image processing algorithms could be used to integrate various radar signals, even if each of them separately couldn't really detect stealth aircraft.


There's no reason why the cameras can't be even better, with some optics added to get performance we need, and put on drones staying in air long time. Sure, drones could be shot down, but drones are extremely cheap and only getting cheaper.

The main problem with cheap consumer drones is that they typically have flight time of only about 20-30mins, so constantly recharging and relaunching them would use a lot of manpower.

So while phone cameras and image processing technology is ready for this today, it might take another decade or two before drone technology catches up.


This approach works far better in daylight than during the night. Night isn't completely dark, so detection might be possible, but it would likely be far more expensive.

It only applies to stealth aircraft, as it flies through open and featureless skies, and it wouldn't work at all with stealth submarines, or even to ground based vehicles since ground is covered with interesting features which could obstruct view or make detection difficult.

Another obvious limitation is that this technology only covers your own country - at least with ground based phones. It won't help you at all if enemy plan is to just show up at your borders, fire some missiles, and head back home.

The whole idea is to have a lot of really cheap short range detectors to defeat stealth, but stealth can still deal with long range detection by radar.

This limitation can be somewhat overcome by deploying camera drones. Drone flying 1km up has horizon at 110km, so it could monitor aircraft approaching your airspace, but that's just a 3 minutes warning, so not exactly amazing.

Another use of this technology could be detecting enemy navy. You'd need to use long flight time drones for that due to distances required, but ships are really big and much slower than aircraft. Then again, surface stealth ships are an extremely marginal thing.

Obviously everything here is back of an envelope calculation. But people's widespread belief that stealth technology is some kind of magic really looks ridiculous to me.