Hacker News Reader

Comment by glbrew | Hacker News Reader

That is fucking incredible they could find that in the Indian Ocean

The Jindalee over-the-horizon backscatter radar in Western Australia has the reach and resolution for detecting a chunk of metal that size on the surface.

Normally it searches for contacts in a programmed grid pattern, sort of a random-chance game of finding bad guys ( HF backscatter radars don't sweep like UHF radars with which we are more familiar ).

I'd hazard a guess that they programmed a radar search of the suspected area and when it found a contact they went and checked the satellite imagery. Spectroscopy from the image capture showed the object to be non-natural.

If this is part of MH370 then you'd wonder why Jindalee didn't pick it up in the air. Just bad luck?

At that time, nobody was looking for it. Unless you know to look out for something specific, it's just one more of the approximately 5,000 aircraft in the air at any one moment. Also these radars can have very narrow beam widths, typically half a degree. Suppose it was already being used for some other task?

But isn't Jindalee primarily an air defence system, intended to look for unidentified and unannounced aircraft?

With no active transponder MH370 would have been unidentified, and no communication with ATC would make it unannounced. It was also unexpected as there was no 'plane due to be in that area according to the filed flight plans (or every other one was identifiable).

Jindalee is marvellous, but unfortunately ( AFAIK ) it isn't possible to use it in a general surveillance manner. It needs to be directed to a specific area which it monitors; this is partly dictated by HF propagation conditions.

Consider that it bounces its signal off the ionosphere to 'illuminate' an area on the surface. You can move that area around like a viewport but you can't make the viewport bigger.

If the MAS a/c didn't enter the areas they were monitoring that day then they wouldn't have contacted.

Presumably the idea is to scan the search area in a methodical (though not completely predictable) fashion, like destroyers trawling the North Atlantic for U-boats?

Yes, I think so! Or set-up a barrier-pattern of searching a linear range of areas continuously. I think they have four or five of these HF radars now, so they could co-operate.

If it's an object that's still floating, it's more likely to be some sort of lightweight composite than metal. And if they are using this radar system in such a way, then they would presumably be able to use it to direct the search teams to the precise location in real time.

Shouldn't this be quite easy to do with existing computer vision technology though (I'd assume this is a computer vision 101 class project)? Especially considering this is the ocean. The images were taken march 16nth, so they probably were running them through their software.

Depends on what you call 'quite easy'. In the terminology of http://fishbowl.pastiche.org/2007/07/17/understanding_engine... , it's still a 'hard' problem. There are tens of thousands if not an order of magnitude more objects and reflections of that size in the ocean.

It certainly is not a 'computer vision 101' project. 'Computer vision 101' is 'do edge detection using opencv', not 'build a complete system that interfaces near-realtime with satellite imagery sources that can do very detailed analysis, but without a real definition of what we're looking for, on huge amounts of data'.

Possibly. We don't really have a dataset of "images of crashed planes" for it to train on though. Perhaps you could use unsupervised learning to find images and image segments that are anomalous. However I expect you'd mostly get pictures of boats, planes, small islands, weird waves, floating garbage, etc. But that does significantly narrow down your search at least.

Funny that the US Military and Intelligence technology was never put to use here. Gives you a sense of the ethics in play.

There's no evidence that they aren't being put to use. The ocean is big, just because one country's assets found something first doesn't mean others haven't been looking.

This area is close to Australia so the Aussie Military is probably better to handle it.

The US is working on the northern tract. Australia + Indonesia got the southern one.

never cited =/= never used

I´m just wild guessing here, but if the US had had the MH370 located with a classified radar or satellite, and they know it had crashed, It would be stupid to disclose that technology. They would better use available resources to point the actual position to the search teams (like: "look at this interesting satellite ping arc"), without showing that they knew the position all the time.

I´m not telling this is the actual case, but I would do it if I were the one in charge of such technology.

People associated with the US government seem to have been giddily feeding the NYT information and theories for days now. If anything the question seems to be how much of what they've been pushing is accurate.

Reply on news.ycombinator.com