jokim@uunet.UU.NET (John H. Kim) (10/03/89)
From: "John H. Kim" <amdcad!jarthur!jokim@uunet.UU.NET> I guess this is the place to ask: In various movies (most recently _The Abyss_) I have seen shots of the inside of submarines which include in the background the person steering the sub. Often, the person is staring at a screen which displays a wireframe image of the surroundings the sub is going through. Presumably, the wireframe is created from a processed sonar signal. My questions: Is this for real? Does it really exist? Where can I get more information on it? Or is it classified? What kind of computing power is required to analyze the sonar info? Thanks in advance. -- John H. Kim | (This space to be filled when I jokim@jarthur.Claremont.EDU | think of something very clever uunet!muddcs!jarthur!jokim | to use as a disclaimer)
steve@uunet.UU.NET (Steve Nuchia) (10/11/89)
From: nuchat!steve@uunet.UU.NET (Steve Nuchia) >From: "John H. Kim" <amdcad!jarthur!jokim@uunet.UU.NET> >steering the sub. Often, the person is staring at a screen which >displays a wireframe image of the surroundings the sub is going >through. Presumably, the wireframe is created from a processed >sonar signal. My questions: Is this for real? Does it really >exist? Where can I get more information on it? Or is it classified? Since I've seen the insides of a real modern sub underway I won't comment on what the Navy does. No research sub would have that kind of gear today. In general sonar data is processed in many different ways to extract many different kinds of information. What I will say here addresses specifically trying to get detailed and accurate information on the topography of the surrounding bottom and the position and shape of other things in the water. >What kind of computing power is required to analyze the sonar >info? Thanks in advance. I can, however, talk about seismic processing, which is so similar to active sonar that almost any analogy drawn will have some validity. The answer to the modified question "how much computing power does it take to make sense of seismic data" is simple -- all you have, and then some. It takes a cray several minutes to figure out a consistent model for a two dimensional slice of dirt from one seismic line, about 100 Meg of (very bulky) data. Such a line would compare with about 30 seconds of sonar data, depending on model and range setting. Extending the solution to three dimensions, which you will have to do to even get it to converge in most cases, takes considerably more power. You aren't going to do it in real time, period. A simpler problem could be solved though. If you restrict your area of interest to a small radius around your sensor, say 100 yards, then refraction efects will be minimal in most cases. It should then be easy enough to pick a few interesting features out of the return, say the dozen nearest planar and point returns and apply something on the scale of a four-node transputer to pinning down the exact location of each. Something like that would be good enough to steer by at low speeds and good enough to direct instruments at very close range. I don't know if anybody has built such a critter though. I have seen reference in the open literature to something called an under-ice sonar or some such name. As far as I can tell it is an analog approximation to this -- a sonar for working close to ice. It should also be possible to work the problem incrementally by interpreting each feature as part of the ground or some moving object. Then you devote some power to keeping your model up to date and some to looking for surprises. I don't know of anybody doing that in real life but it is the kind of thing being applied in robotics research. -- Steve Nuchia South Coast Computing Services uunet!nuchat!steve POB 270249 Houston, Texas 77277 (713) 964 2462 Consultation & Systems, Support for PD Software.