What It Would Really Take To Sink a modern Aircraft Carrier

[jhungary]

@jhungary
I think it is quite easy to detect 3d objects from a 2d image. Nimitz class carrier is 332m approx, surely I can get a good enough 3d model of it from net. I have my satellite's attitude angles with me, using that I can do a 3D to 2D projection of the carrier's top from my that particular angle and superimpose on the images I receive from the satellite. Size of 332m, it can be easily detected by mil satellites, for a good enough match of say 85% you can call a human. If it matches, the satellite has only to track in the vicinity of the previous image, knowing the time stamp and max speed of carrier it would be done.
What's wrong with this?
This won't take time once a match is made. The only lag will be from satellites transmission and whether there are other satellites covering that particular area.

Key word, Approx.

You cannot ask a computer to put a positive result if you simply guess the dimension. Is it 332meter? or 331 meter? Or 333 meters? Or even is it 332.11 meters or 332.12 meters?

To be able to make a 3-D Projection, the dimension MUST BE CORRECT. Which mean if you have a matrix that serve 332.11 meters, but the actual dimension are calculated as 332.12 meters, your computer will give you a negative result.

The thing is, nobody actually know exactly how long a carrier is, the only way you can make a 3-D projection is to actually scan the carrier down to the last millimetres and compare it to the image you obtain via satellite, but I don't think anyone will give you the actual dimension of the ship. then you failed.

In the car's example, if you can get a particular signature, say licence plate, you can track it? Just need good processors.

The problem is, license plate are unique (the number is unique) If you can read the license plate, that's the ultimate reference point of an object. On the other hand, Carrier and other shipping does not carry a plate with unique number to identify them, (for example, pennant 68 in USN stock could be CVN-68 USS Nimitz, could be USS Anzio CG-68, could be DDG-68 USS The Sullivan, Could be actually any ship with an identifier 68.)

Do you know how horizon detection works? The earlier ones used rgb difference between that of sky and ground to detect the horizons. You can also use texture to differentiate between the sky and ground.
When I have an image from the satellite, I can do a simple rgb analysis to point the odd ones out. In the image of an ocean, it will be a vast expanse of blue. Any ship will stand out on rough rgb analysis. Using this odd shape out and comparing with 2D projection of carrier knowing the attitude angles of satellites, I can compare the two and get a match. No requirements of landmarks.
Every problem has got its own solution, here you don't need landmarks.

You are talking about contrast, not reference point.

The image detection is always on vectoring detection, you have to have an unique identifier that's either a set of unique state (distant reference between a given set of 2 points) or a unique identifier (eg. a special color scheme)

And again, you cannot compare the 2-D projection of the carrier if you do not have the actual dimension.

 

[Taj_91]

Key word, Approx.

You cannot ask a computer to put a positive result if you simply guess the dimension. Is it 332meter? or 331 meter? Or 333 meters? Or even is it 332.11 meters or 332.12 meters?

To be able to make a 3-D Projection, the dimension MUST BE CORRECT. Which mean if you have a matrix that serve 332.11 meters, but the actual dimension are calculated as 332.12 meters, your computer will give you a negative result.

The thing is, nobody actually know exactly how long a carrier is, the only way you can make a 3-D projection is to actually scan the carrier down to the last millimetres and compare it to the image you obtain via satellite, but I don't think anyone will give you the actual dimension of the ship. then you failed.



The problem is, license plate are unique (the number is unique) If you can read the license plate, that's the ultimate reference point of an object. On the other hand, Carrier and other shipping does not carry a plate with unique number to identify them, (for example, pennant 68 in USN stock could be CVN-68 USS Nimitz, could be USS Anzio CG-68, could be DDG-68 USS The Sullivan, Could be actually any ship with an identifier 68.)

I will start my work with 332m, do a 2D projection and try to get a match. I am not looking at 100% match. A good enough probability will make the alto call a human.
And there is no shortage of time, I can start my track once the carrier has left its port.
For a carrier, my signature would be its unique shape, the external ones.

 

[jhungary]

I will start my work with 332m, do a 2D projection and try to get a match. I am not looking at 100% match. A good enough probability will make the alto call a human.
And there is no shortage of time, I can start my track once the carrier has left its port.
For a carrier, my signature would be its unique shape, the external ones.

How a carrier is a unique shape?

One simple problem, aircraft and people constantly walking around and pushing around on the deck of the carrier

For example.

Both are USS Abraham Lincoln, however, the way the deck arrange is different, hence, in computers term, it is a different picture. But if I ask the computer if this is the same ship, judging by pixel analysis, the computer will say no.

You are talking about multiple possibility here. Abe with 1 aircraft on deck and Abe with 2 aircraft in deck is a different states, Abe with 3,4,5,6,7,8,9,10......79 aircraft is all different state, then you have the different arrangement of the aircraft. Different aircraft protrude outside the ship boundary will extend the ships dimension in variable, all these will read differently when the computer analysis the pixel between the two. How are you going to define all these exception??

 

[Taj_91]

You are talking about contrast, not reference point.

The image detection is always on vectoring detection, you have to have an unique identifier that's either a set of unique state (distant reference between a given set of 2 points) or a unique identifier (eg. a special color scheme)
And again, you cannot compare the 2-D projection of the carrier if you do not have the actual dimension.

How a carrier is a unique shape?

One simple problem, aircraft and people constantly walking around and pushing around on the deck of the carrier

For example.

View attachment 392960

View attachment 392961

Both are USS Abraham Lincoln, however, the way the deck arrange is different, hence, in computers term, it is a different picture. But if I ask the computer if this is the same ship, judging by pixel analysis, the computer will say no.

You are talking about multiple possibility here. Abe with 1 aircraft on deck and Abe with 2 aircraft in deck is a different states, Abe with 3,4,5,6,7,8,9,10......78=9 aircraft is all different state, then you have the different arrangement of the aircraft. Different aircraft protrude outside the ship boundary will extend the ships dimension in variable, all these will read differently when the computer analysis the pixel between the two. How are you going to define all these exception??

I am talking about the silhouette of the ship. I am not bothered with what is on top of it.
There is stark difference between the silhouette of carrier ship to that of cargo ship:

Compare this with silhouette of carrier.
Had it been a ww2 carrier like this, I will have trouble

 

[jhungary]

I am talking about the silhouette of the ship. I am not bothered with what is on top of it.
There is stark difference between the silhouette of carrier ship to that of cargo ship:
View attachment 392964
Compare this with silhouette of carrier.

lol, you still don't understand.

If you do that, all it will tell you is that the picture is not a cargo ship, it will not tell you it is an Aircraft Carrier, let alone American Aircraft carrier.

I don't know if you realise this, but what you are saying is only half of the comparison process, to compare a silhouette, you need to extract the silhouette and compare it to SOMETHING.

As I said, fighter aircraft protrude outside the ship, which alter the silhouette of the same ship. If you compare the 2 picture of the Abraham Lincoln in my last post, the two ship will have a different silhouette, if you are using one of them as a base point to a Nimitz Class carrier, then it will tell you this is not an Nimitz Class carrier. As the ship silhouette have been extended by the F-18s and E-2C differently.

 

[Taj_91]

lol, you still don't understand.

If you do that, all it will tell you is that the picture is not a cargo ship, it will not tell you it is an Aircraft Carrier, let alone American Aircraft carrier.

I don't know if you realise this, but what you are saying is only half of the comparison process, to compare a silhouette, you need to extract the silhouette and compare it to SOMETHING.

As I said, fighter aircraft protrude outside the ship, which alter the silhouette of the same ship. If you compare the 2 picture of the Abraham Lincoln in my last post, the two ship will have a different silhouette, if you are using one of them as a base point to a Nimitz Class carrier, then it will tell you this is not an Nimitz Class carrier. As the ship silhouette have been extended by the F-18s and E-2C differently.

I am comparing it with 2D projection of the carrier that I already have. I have the distance, the attitude angles of the satellite, I will do the projection and compare. Also, I don't need to know the exact dimensions to make the projection, I just need good enough match.
Yes, the protruding aircrafts will change the silhouette, but there image processing can be done. It will depend on the size of the image from the satellite, whether the satellite is geostationary (huge distance) or low earth orbit (small distance). Processing tools are advanced too.
I haven't tried it of course, but I think the image can be processed. Up to you to disagree. Might be difficult but not certainly impossible if you spend a couple of million dollars,

 

[KAL-EL]

A few are talking as if this is only limited to US CBGs that could be targeted and attacked

Current/future CBGs from other nations could be targeted/attacked.

Might be wise for a few of the PDF hyper nationalists who live in a video game fantasy bubble to remember that. It could just as easily go both ways.

But besides that. A very fascinating rational discussion/debate for the most part gentlemen. kudos

 

[jhungary]

I am comparing it with 2D projection of the carrier that I already have. I have the distance, the attitude angles of the satellite, I will do the projection and compare. Also, I don't need to know the exact dimensions to make the projection, I just need good enough match.
Yes, the protruding aircrafts will change the silhouette, but there image processing can be done. It will depend on the size of the image from the satellite, whether the satellite is geostationary (huge distance) or low earth orbit (small distance). Processing tools are advanced too.
I haven't tried it of course, but I think the image can be processed. Up to you to disagree. Might be difficult but not certainly impossible if you spend a couple of million dollars,

............

Let me save you some time, someone from Raytheon had already tried on my unit's Satellite Station when I was still in the Military, the result is negative.

I don't know what kind of sample you use or what kind of algorithm you are using, but unless you can use an unbiased random based sample, however you try will not be accurate

 

[Taj_91]

............

Let me save you some time, someone from Raytheon had already tried on my unit's Satellite Station when I was still in the Military, the result is negative.

I don't know what kind of sample you use or what kind of algorithm you are using, but unless you can use an unbiased random based sample, however you try will not be accurate

Haven't tried it yet, but was just discussing possibility. But in military everyone plans for the worst, maybe the Chinese have developed it? Its a good possibility, not bad one!

 

[jhungary]

Haven't tried it yet, but was just discussing possibility. But in military everyone plans for the worst, maybe the Chinese have developed it? Its a good possibility, not bad one!

We did plane for the worse, that why the US government is trying it. That is the reason I know about it

I can also tell you in some degree, what we are talking about is tried on the famous PRISM, but as to how much and how far it goes, that I cannot tell you.

 

[Mugwop]

If a submarine can stay at a depth of more than a 100 feet and sneak up on a ACC then all it takes are small torpedoes which can be hard for cat systems to detect.

 

[gambit]

I am talking about the silhouette of the ship. I am not bothered with what is on top of it.
There is stark difference between the silhouette of carrier ship to that of cargo ship:
View attachment 392964
Compare this with silhouette of carrier.
Had it been a ww2 carrier like this, I will have trouble
View attachment 392965

You should. Or rather -- MUST.

Whatever is your sensor, radar or photo, in order to distinguish the unique outlines of container transport vs aircraft carrier, your sensor is sensitive enough to distinguish out the details that are on the surface of the ships, even in a 2D view. But then, this is why SAR capability is enabled on the next generation of radars installing in the non-5th gen fighters.

But if you chose not to distinguish out those unique shapes of each type of ships, then you can and will be fooled.

 

[Oldman1]

As if SM-3 has proven interception against a dozen missile saturated attack. US are hapless against mass of DF-21D attack. Even 24 DF-21D needed to sunk one USN CVN is dirt cheap in terms of cost ratio.

LOL! With the amount of ships surrounding the carrier, it would take more than 24 to try to sink the carrier. And haven't even mentioned the SM-6 that can intercept during terminal phase besides the SM-3 which intercepts mid course. And in the near future there is possible BMD ship that can handle larger salvos of anti ship ballistic missiles. So imagine building about a dozen for the carrier groups.

Sir I agree that both systems are unproven in war scenario and I also accept that USA experience and tech is way too superior but to hit a bullet with a bullet is way too tough ... So the job of DF21D is difficult but the job of anti-ballistic missile or SAM system is even more difficult especially against saturation attack ... Furthermore, DF21D has to hit just once whereas SM-3 has to be successful in all the cases therefore even in terms of probability DF21D job is easier ...

Actually the DF21's job is just as difficult if it cannot detect its target. You depend on satellites to detect the ships. What if they are destroyed or jammed? AWACs or drones won't help because they be detected long before getting even anywhere near the proximity of the carrier.

 

[Taj_91]

We did plane for the worse, that why the US government is trying it. That is the reason I know about it

I can also tell you in some degree, what we are talking about is tried on the famous PRISM, but as to how much and how far it goes, that I cannot tell you.

Prism? It was data mining programme or military data analysis (carrier detection) programme?

@gambit I am taking the shape of ships into account, but not the cluster at the top. BTW, I was just trying to show that detection is not an impossible task, can be done. Uncle Sam himself might be doing it, lol!

 

[jhungary]

Prism? It was data mining programme or military data analysis (carrier detection) programme?

@gambit I am taking the shape of ships into account, but not the cluster at the top. BTW, I was just trying to show that detection is not an impossible task, can be done. Uncle Sam himself might be doing it, lol!

image recognition, not carrier detection is used for PRISM. In order for PRISM to run in automation, you have to have a way to ID a target by image recognition.

i.e. if someone send someone a photo over the internet and that someone is being monitored by PRISM, you will need to know who was in that photo to be flagged, to be autonomous, they will need a image recognition software suite to match the photo that someone send to the terrorist database, you cannot ask that someone to take a photo at the same distant and the same angle as with the one you had in the database, so a geometrical match is impossible, you must have image recognition.