Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
Technology

3-D Surveillance Technology 93

Posted by CmdrTaco
from the i-choose-the-3d-pill dept.
scubacuda writes "According to this Technology Review article, a new surveillance technology called Video Flashlight melds 3-D models from background scenes. This "tweening process" allows security persononnel to fly around a subject such as a pedestrian, getting a detailed look without jumping between widely separated views." That's just flat out cool.
This discussion has been archived. No new comments can be posted.

3-D Surveillance Technology

Comments Filter:
  • This is a direct descendant of the technique used in the Matrix.

    From the rumormill, this technique will be used in the Matrix 2 to create even better effects than was done previously.
    • When I first saw The Matrix, I thought they did the opening scene's jump-kick with tweening, and thought it was too cool. Then I saw The Making of The Matrix, and that they did it with the multiple computer-timed still cameras, and was not as impressed as I was before.

      If the 2nd movie uses tweening, and the result is better looking than the timed cameras used in the first, I will be impressed again.
      • > If the 2nd movie uses tweening, and the result is better looking than the timed cameras used in
        > the first, I will be impressed again.

        Uh, how exactly can an interpolated, computer generated image of a human be better looking than an actual photograph, which is what M1 used? It might buy you more flexibility and more otherwise impossible scenes, but it's certainly not going to be looking more realistic than the real thing.

    • by clifyt (11768)
      Hmmm...Probably not. This technique has been around for a while, but not really practical or smooth except with recent technologies. The Matrix trick was very unsophistocated in terms of electronics. Its just a few hundred cameras that are synched and can be switched from in a linear method giving the illusion of 1 single camera.

      Several years ago, there was a product released on the Mac (and then simply disappearing) called Canomera. It took 2 camera positions and after you filled in a few details such as corners of buildings and all that stuff, it could interpolate between the 2 (or more) picts and allow this same sort of trick. If one had static cameras with these corner positions prelocated, it would be dead simple to create a realtime fly through with the same software base.

      clif
      • I agree. The multiple-camera technique has been around since the mid-90s, AFAIK. First used in MTV videos, I think.

        The method we are talking about sounds more like the method portrayed in Enemy of the State, as mentioned above. I saw an interview with the director, and he said everything mentioned in the movie comes from *declassified* information-- in other words "old news".

        Yes, the tweening techniques have been around for awhile, too. I remember seeing at least 5 years ago a documentary on a project that combined a digital camera with GPS positioning data (and a gyroscope, I believe) to create a 3D model of a landscape. They showed a guy just walking around a park with a video camera and a laptop; he didn't even have to walk in a predermined path. NASA used this same technique for the 3D model of the Mars Rover landscape.
      • Its just a few hundred cameras that are synched and can be switched from in a linear method giving the illusion of 1 single camera.

        Actually no. There would have had to have too many cameras to get a smooth camera movement- and the cameras would have got in each others way. The Matrix effects really did have to do inbetweening between the different cameras digitally in fact.

        I am uncertain as to how much of this was done by software and how much was hand hacked in practice, frame by frame; but it wasn't as simplistic as you said above.

        • Actually, yes. See http://www.howstuffworks.com/question295.htm

          If you see the dcumentary of the making of "The Matrix", you can see the actual setup. It doesn't take hundreds of cameras, though. Film operates at 24 frames per second, so 2 seconds worth of "fly around" would only need 48 cameras, give or take.

          I'm sure it would be possible to use tweening to assist in the motion, but hardly necessary. This effect has been around for a few years, definitely before video and filmmaking went completely digital.

          • Did you even read the url you reference?:

            "A technician deals with all of these imperfections one image at a time using a computer and digitized versions of the images. Once the still images are perfect, the morphing software interpolates between them. Then the background images are laid into the green area. A technician has to build a complete 3-D computer model of the computer-generated scene and then key the rotation through this scene to the position of the camera in each frame of the film."

            As I said, the cameras are too bulky... they needed to be able to inbetween them to get the effect that the director wanted.

  • If you can't see all of someone how the hell are you supposed to get a 3d model of them! Admitedly you could assume that their symetrical or something, but that is usually a bad assumption. And who is to say that the dimension you can't see isn't normal. Like a person who is 5'7" and three feet wide but whose depth is normal. Their side veiw would imply a normal 5'7" person.
    • by telstar (236404)
      You "DO" see all of someone. As the article says, this technology just uses hundreds of cameras, and merges their information in realtime. Of course the criminal might notice the couple hundred cameras as he walks down the street .... but at least you'll have a 3D view of him as he steals one of them.
    • I really wish someone who knew more about this would post, my super charged troglodyte mind tells me it works something like this.

      you'll need at least 6 cameras:
      the scene from above:

      /**c**\
      /*-----*\
      /**|+++|**\
      c**|*c*|***c
      \**-----**/
      \*******/
      \**c**/

      in this crude ascii drawing we have five cameras(c), there's another one below the middle one, so we could say 6. there the box in the middle is the scene, and the diagonal lines is the cameras viewing frustum. the (*) is just for spacing, i couldn't make it work with spaces.

      Each one actually contains two cameras, very close to each other to create a stereo effect, which presumably some software to extrapolate to create a 3d image.

      The 3d image from any one camera would only be good for what it can see, so you'll have an extrusion, or unknown for what it can't. this is there the other cameras fall into place, they each can generate an image for what they know. Then though math and magic all the 3d images are combined to create one 3d image.

      this is only a guess, maybe it works like this maybe not.

      -Jon
    • by Anonymous Coward
      Actually, they talked about multiple networked cameras.
      It really doesn't matter how many cameras you have, as long as you can see every part of the subject, which means a minimum of three cameras.
      As they explained in the article, the basic buildings in the area have already been modelled in 3D and are used as reference for the computer to be able to discern what is what. Anything which doesn't match the model, is considered a foreign object, and by extrapolating information about that object from as many cameras as it needs to, the computer can build a 3D model of that object in real time. Since this is being fed into a 3D engine, the observer can place a virtual 'camera' like he would in any 3D animation package, anywhere within the scene. Of course, this will basically mean the end of privacy. Imagine going to a romantic dinner, and the security guy on the 3rd floor is looking at your date from the same angle you are... It's all possible.
      Of course, Hollywood will take massive advantages of this technology, using it to generate camera moves in a real location otherwise impossible to achieve.
      Also, imagine watching a ball game on TV where they can cut to any angle, even seeing the POV of the ball as it travels through the air.
      Or, being able to see the exact POV of a race car driver without puttin a camera on his helmet.

      Next stop: displaying the output in 3D using holograms. That could be very cool for ballgames also... Your home team is on the road, but you can still go see them 'live' at their home court.

      Oh, and pr0n will take on a whole new meaning...
      • Oh yeah.. it's the end of privacy.
        Hello.. if a camera can't see anything at any given moment, neither can the security guys.

        So if your date is sitting around, they *can't* see it from your point of view accurately, because you would be obscuring the camera. What? The camera is above your head looking at your date? then you never had privacy in the first place.

        The hologram thing is cool.. someone just has to invent the hologram

    • This isn't about precise 3d imaging of each person wandering around.. it's about a model that's easier to work with.

      Following people around using discrete cameras is not natural... it's difficult.

      Using those cameras to create a more natural 3d environment for those monitoring to see what's going on will make it easier for security staff to have a good feel for what's up.
  • I remember a court case where some politician was reported in the paper to have been with a prostitute.
    Later on some late night talk show they got footage of him just walking along to some one and talking to them.
    Then they used software like this to place him on a different background going over to a prostitute.

    It was so convincing he took them to court for making it.
  • pr0n (Score:4, Funny)

    by Max von H. (19283) on Saturday June 22, 2002 @10:25AM (#3749069) Homepage
    I can foresee a bright future for this technology in the pr0n industry... Combine it with a VR helmet, and you're on the way to re-enact some cool scenes from Strange Days!
  • Scary... (Score:1, Flamebait)

    by donnacha (161610)


    It's kinda scary that the reporter is, in a sense, selling us on this aspect of the ever more frightening reach of surveillance into our lives by tying it in so heavily with one of the coolest films of recent years, The Matrix.

    While the film itself was pretty damn cool, we should bear in mind that the world it depicted would be pretty damn shit to actually live in.

    • While the film itself was pretty damn cool, we should bear in mind that the world it depicted would be pretty damn shit to actually live in.

      Only if you knew what's going on. It's an exact simulation of real life - you'd have no idea if you were in it. It wouldn't be "shit to actually live in" - it'd be the same as our world.
      • It wouldn't be "shit to actually live in" - it'd be the same as our world.
        No.

        It wouldn't be "the same", it would be pretending to be the same as our world.

        It's like when you've got a friend who's living in a complete fantasy world: he might be perfectly happy right now but, as his friend, you have a duty to at least try making him see reality because if he realizes it for himself years later, he'll bitterly regret pissing away valuable years of his life working for and believing in something that actually wasn't at all what he thought it was.

        One of my friends used to work for Microsoft.

        • If this fantasy world is convincing enough for everyone to believe it, I don't really have a problem with it. Remember Cypher - he'd rather live in the Matrix than in the real world, and I suspect most people would feel the same. Steak vs. goop? Hmm... I think I'll live in fantasy world, thanks.
          • Steak vs. goop? Hmm... I think I'll live in fantasy world, thanks.
            If life was just a question of consumption, of watching the pretty colors go by, sure.

            But perhaps it's about leaving your mark on the world, the real world. Cows might be happy but, essentially, we've redirected them from their destiny to serve our ends.

            I'm not saying that there is a specific point to life but, getting very minimalistic about it, you could say that our "purpose" is to pass on our genes in the ongoing dance of evolution. If we are not at least partially in control of our destinies we no longer get to participate in that, the continuance of our lines.

            It's not so much a question of Steak vs. Goop but what you do with your goop.

      • There are times when this world can be shit to live in too... :-)
        • There are times when this world can be shit to live in too... :-)
          Yeah, but at least it's real and, assisted by a reasonably accurate idea of it's nature, we are free to do our best to improve it.
    • Was anyone else bothered by the phrase in the article "setting off an alarm when an unusual pattern is detected"? Great, now we'll have to practice "acting normal" to avoid triggering an alarm. Since it's legal to install cameras in public places, you could potentially be monitored almost anywhere. Hmm, this reminds me of a certain novel by George Orwell (not the one with the pigs.)
      • Since it's legal to install cameras in public places, you could potentially be monitored almost anywhere. Hmm, this reminds me of a certain novel by George Orwell (not the one with the pigs.)
        Yeah, well, something tells me there's going to be plenty of pigs involved with this technology.
    • by grokk (227023)

      The incessant augmentation of police state powers is NOT a cool thing at all.

      The way a lot of people around /. talk about these things, it's pretty clear that they don't ever expect to be the object of these new 'toys'. I find the complacent, pseudo-cool, abstract discussion of these matters to be almost as scary as this police state 'apparatus'.

      Tell me people: just when do the warning bells go off in your heads (the 'crime' issue is always meant to sidetrack your critical reasoning powers, BTW)?

  • by allism (457899) <alice.harrison@gm a i l .com> on Saturday June 22, 2002 @10:31AM (#3749074) Journal
    Yeah, I know they say that they are using multiple cameras and just using tweening to fill in gaps between frames, but when you start sticking little stuff in, how do you know what's real and what's not?

    Made me think of the character in Stranger in a Strange Land, can't remember her name, but she was a trained witness of some sort and would not testify, for example that a house was white on all sides unless she had seen all sides of the house and then she wouldn't testify that the sides of the house she could no longer see had stayed white. Granted, that's a little extreme, but the average Joe is gonna believe what he sees is true, and this is mucking with that truth just a little bit. If this is used in court, is an expert going to be able to testify that only certain parts were added? Wouldn't that seem a little odd to a jury, "Hi, we're going to show you a videotape of something that happened and you're going to have to take it as gospel...AFTER we tell you that we added a few little bits to it."?

    Course, I'm paranoid, if someone tells me they mucked with it a little bit, I'm gonna assume they added exactly what they wanted me to see...
    • by ceejayoz (567949) <cj@ceejayoz.com> on Saturday June 22, 2002 @10:35AM (#3749091) Homepage Journal
      If they showed the tapes in court, they'd be able to show the original, uncombined versions too. Remember, it's not some magical camera - there're many many cameras that get their input processed for easier viewing. You can still fall back on the raw data if you need it.
    • I suppose that depends on how much the technology is (or becomes) trusted.

      Yes, it's possible to "add" a gun (or a prostitute) to a scene. It would also be possible to add a digital watermark to each source frame and reconstruct the tweened data (or what the surveillors saw) on the fly for legal purposes (assuming, of course, that some enterprising soul didn't find a way to hack the watermarking scheme).

      The technology sounds cool though. Imagine how cool pr0n dvd's could be. You already have multiple camera angles available on disk, all you'd have to do is to play back the disk on a tween-enabled player or a computer and get a whole new perspective on things. yowsa!
    • There is a much bigger issue here. Digital images should not be admissible in court. Anyone with a simple image editor can add/remove things. Photographs at least have a negative that will show any tampering.

      Until digital cameras "sign" every image with a unique id, a time stamp and a hash of the image don't expect the courts to accept the images as evidence. The camera should also have some way of verifying that it took the photo image and that it hasn't been altered.

      Digital video would have to have every frame signed. Once you can verify that the original video sources are free of tampering then there shouldn't be any problem with showing a 3D fly through providing the court can locate the original view that contributed to some item of interest.

    • When the movie "Outbreak" and its CG F/X with choppers hit the screens, I remember thinking "Wow, you can't tell what's real and what's not anymore. You really cannot tell the difference between a CG scene and a real one. I wonder how long it will take before news studios start using this to fabricate stories."

      Upon which a friend of mine replied, "What makes you think they don't already?"
  • what's the deal with the word 'tweening' ?

    is interpolation too big of a word for the average person? why does the word 'tweening exist' ?

    i hate the word.
    • Re:tweening (Score:2, Informative)

      by Nurgster (320198)
      Tweening is just a specific type of interpolation (in graphics, anyway).

      It involings interpolation of vector coordinates, like morphing is interpolcation of bitmaps.
  • cause we all know that the security guys viewing this would be looking at the woman in the red dress.
  • How any new technology designed for surveillance can be described as "flat out cool".
  • It's a neat idea, but I don't think it will be as useful as one might hope. Anything in the computer image would still have to have a direct line of sight to a camera. It sounds like a great tool, but remember computers can't do *everything*
  • This has already been in use for years!
    Its been featured in countless movies, along with the systems that can "enhance" fine detail out of 4 pixels of NTSC video.
  • by krb (15012)
    Didn't they use this feature in the super bowl a couple of years ago? Of course, you still need one camera to expose each angles, usually at least 3, and a hoss computer to build the 3D model. If i recall, it wasn't as Matrix-esqe as we'd like it to have been, (though that's mostly a limitation of camera and computing power).

    Aside from all that, what's the point. If a guy looks suspicious, and you have 3 cameras able to pick him up, flip the camera view -- is that too much harder than rolling a jog dial?
  • by FFFish (7567) on Saturday June 22, 2002 @11:13AM (#3749200) Homepage
    First, a sensible editorial [privcom.gc.ca] from the Vancouver Sun.

    Second, some words from the Canadian privacy commissioner [privcom.gc.ca], in which he comes down on video surveillance.

    Third, the cameras are ruled illegal [oipcbc.org].

    Canada has a privacy commissioner who is independent of the government and police and who has one overriding concern above all else: ensuring that the constitutional privacy rights of the Canadian public are respected.

    In the past, he's also prevented the government from creating a super database that merges all information from all sources -- police, medical, political, taxation, etc -- into one system. So ruled because it would make it far too easy for the various branches of government to look at data they shouldn't have access to.

    Thank goodness Canada's got the foresight and commonsense to have an independent commissioner!
  • Isn't this like that they were using in Enemy of The State to try to see what was dropped in Will Smith's bag by that guy that was later killed on the road?

  • Courts? (Score:1, Troll)

    by ZaneMcAuley (266747)
    This "tweening process"... hmmmmmm

    Computer added images, surely thats not CONCLUSIVE PROOF? as the image was rendered by the computer and not ACTUALLY there? Basically its an educated guess, not fact?

    Could this stand up? or be questioned?
    • If they're going to be recording sections of the video, they'd do so from the raw stream, not the omputer generated 3D environment. This is just a better way of representing multiple camera shots of the same scene.
    • Conclusive proof of what?

      The tweening is so the image looks smoother and more realistic, not to change what happened.

      The matrix used tweening to calculate the images between images to make things look smoother and more pleasant instead of jerky.

      Yes, it's an educated guess.. but the tweening doesn't place a guy in a room when he's not there, or make him open up a door that he never really touched.. it just makes it look better and more natural.
  • I'm sorry, but any footage that is interpolated or derived from existing frames cannot be admissible in court. It should be considered as speculation. If it isn't then we have some serious problems.

    I'm all for the added benefits to pr0n, though.... :)
  • What a bunch of misleading hype! This article tries to scare all the Chicken Littles out there into beleiving that once again the sky is falling. (Remember the last time the sky fell on Jan 1,Y2K) The story insinuates that "big brother" will soon be counting the hairs in our nostrils by simply interpolating the images picked up by a few traffic cameras placed around town. The key here is that it takes HUNDREDS of cameras to perform this feat and these cameras need to be networked together at high speed before being processed or "tweened" if you prefer. An istallation of this sort would run a round $1 million bucks per city block for the initial installation alone. 'Guess were safe for now. Don't get me wrong, I love the technology, I just feel bad for the Chicken Littles who will beleive it and then never leave their (hen) house anymore.
  • From the last paragraph of the article:

    But early versions have already been installed at U.S. Army Intelligence headquarters and are under consideration for New York City's three airports--perhaps bringing us all a step closer to living inside the Matrix.

    If by "living inside the Matrix" you mean living in a world where our every action is monitored by technology then, yes, we are getting closer to that.

    I don't doubt that this thing has legitimate uses, but I'm not about to jump up and praise every single development in survelance technology. I think Big Brother's eyes are allready a little too sharp, and I'd rather see them make improvements on how they apply their technology. There was no technical reason for intellegence to miss plans for Sept. 11th, and no array of videocameras etc. would have helped. The problems were organizational, and I don't think sufficient improvements have been made in that area.

    Bottom line: you get watched, the terrorists go unnoticed.

    • The problem with big brother isn't his eyes, it's his inability to stay the fuck out of your business when you're not hurting anyone. I'm not ashamed of anything I do, but some of it may very well be illegal or considered immoral by people who have the power and inclination to stop me and punish me. But if people were actually only punished for things which hurt other people, and the system were designed to work with such fair overseers...

      heh. heh.

      Anyway, IF those things were true then it wouldn't matter who saw you do what, now would it? The problem isn't what people see, it's what they do about it.

      As a side note; I believe in privacy, but if you were never given any reason to believe that you had been seen - any reason - then it would be like not being watched. Unfortunately, it just doesn't work that way - But the ObDisclaimer snuck in there.


    • For Christ sake ... don't you know you are a battery being consumed by the government ? We are one of many in the geowolf cluster of human intellect.
  • Let's face it. Security is a friggin' boring job. I know people who have done it. About the only thing that keeps it from driving the people doing it completely insane with boredom is the benefit of zooming in the cameras on women with lots of cleavage showing.

    So you can imagine how this software will be abused.
  • by Thagg (9904) <thadbeier@gmail.com> on Saturday June 22, 2002 @01:09PM (#3749555) Journal
    Check out this paper by Chen and Williams. In this work done back at Apple in '93 they describe how to create intermediate camera angles from multiple static images. [nec.com]

    Of course, the capacity to fly around the scene in real time had to wait until computers got a lot faster.

    thad
  • Spectator mode? (Score:1, Interesting)

    by Anonymous Coward
    Sounds like they want to get spectator mode like in Counter Strike. Right-click to jump between people walking around, and you can move the mouse to change the angle you're looking at him with...

    At least that what it sounds like to me...
  • I'm surprised nobody else has compared this to the "Esper" device used in Blade Runner, which allowed Deckard to "see around corners" in static images. Though, to be fair, I can't think of any *possible* way this could be done using only one image - unless the machine was extrapolating from extremely subtle shadows and reflections on other objects on the picture. Even then you'd get a very crude image of unseen objects at best.
  • I remember a scene from Enemy of the State where the evil NSA geek takes a shot of Will Smith's shopping bag captured from a lingerie store security camera, and rotates it in 3D, filling in information as he goes along. I was like, "this is soo much bullshit. Typical Jerry Bruckheimer film - junk science all the way."

    Well, I guess it wasn't totally bullshit. However, if I find out that the DNA pattern of an ideal brunette can be modeled using only one package of M&Ms, I'll have to shoot myself...
  • sure 1984 passed. but we are getting closer to it.
  • I thought it was pretty neat when a football program would do replays of a scene using multiple cameras and "swing" you around a stadium to a better viewing angle. I think it was Fox Sports that did this, but I'm not sure.

    Another thought: One can estimate that there are three or four phone numbers for every man woman and child (home/work/cell/fax/modem/etc.). How many cameras per person will we have within 10 years as we move to a surveillance-oriented society?

    -ez
  • by ap0 (587424)
    you're all paranoid about being watched, aren't you?
  • by mortal (2462) on Saturday June 22, 2002 @05:42PM (#3750525) Homepage
    The Sarnoff corporation has more information available on their homepage, including a downloadable video clip of the the flashlight in action; available here [sarnoff.com].
  • This could have some really awesome uses in Teleimmersion and Virtual Reality.

    Think about the last videoconference you attended: it was probably a far cry from being face to face. Adding simple stereo vision to that probably would not do much for interacting with your peers. However, a system like the following might change things.

    Here's what I'd propose, if I could build my "dream system":

    Set up the following:
    *Two conference rooms equipped with the cameras mentioned in the article
    *Optical See-through HMDs [unc.edu] that the users would wear
    *A very fast network connection [internet2.edu] between the two locations
    *Software to make it work.

    This way, individuals in disparate locations could walk around, talk to each other, and do everything but shake hands. Actually, get one of these [mindflux.com.au] and that might even be possible :)

    You would also somehow do a similar trick with the audio to enable "private" conversations between individuals sitting next to each other.

    This could be the first really useful immersive application. Think about how much travel time would be saved.

    Any thoughts?
  • Yep, in that movie they use this 3D tech to pan around a shopping bag to see what it contains. But this would never work in reality...
    • Really? why not?
      The camera saw the bag from several angles over time as the guy walked away.

      You then use a computer to model what the bag should look like from whatever angle. You aren't displaying anything you don't already know.

We will have solar energy as soon as the utility companies solve one technical problem -- how to run a sunbeam through a meter.

Working...