Slashdot Mirror

← Back to Stories (view on slashdot.org)

Seeing Buildings Shake With Software

Posted by samzenpus on from the whole-lot-of-shaking-going-on dept.

mikejuk writes: In 2012 a team from MIT CSAIL discovered that you could get motion magnification by applying filtering algorithms to the color changes of individual pixels. The method didn't track movement directly, but instead used the color changes that result from the movement. Now another MIT team has attempted to put the technique to use in monitoring structures — to directly see the vibrations in buildings, bridges and other constructions. Currently such monitoring involves instrumenting the building with accelerometers. This is expensive and doesn't generally give a complete "picture" of what is happening to the building. It would be much simpler to point a video camera at the building and use motion magnification software to really see the vibrations and this is exactly what the team is trying out. Yes you can see the building move — in real time — and it seems to be a good match to what traditional monitoring methods say is happening. The next stage is to use the method to monitor MIT's Green Building, the Zakim Bridge and the John Hancock Tower in Boston.

21 comments

  1. camera shake? by calzones · · Score: 3, Interesting

    Accurately and precisely canceling out camera shake (indeed the movement of the building/foundation where the camera itself sits too) on something like this would seem to be a big deal.

    I guess instead of having accelerometers on the building they put them on the camera? Article didn't really get into this aspect.

    --
    Asking people to think is like asking them to buy you a new car
    1. Re:camera shake? by Anonymous Coward · · Score: 2

      The camera's own motion should produce an identical effect across the field of vision, for which the software should be able to control. The motion of objects in view, however, will differ relative to other objects in the field, and the algorithm likely checks for difference between subpixel color change in different regions of the image.

    2. Re:camera shake? by Anonymous Coward · · Score: 1

      It works on *colour* changes in the building (because og lighting reacting to the motion I presume). Your local camera shake is not going to change the lighting on the building.

    3. Re:camera shake? by Anonymous Coward · · Score: 0

      It's the opposite of video stabalization with the addition of a bandpass filter.

      Imagine DSP shakey-cam that only exaggerates/distorts/embosses the video around certain frequencies of motion.

      OpenCV has a function called absdiff(inmat, outmat); that gets the raw motion data and then they use the pixel motion as a seed for a bandpass filter.

      I'm not sure if their algorithm uses dense optical flow or not. You'll notice that absdiff does edge-detection(as a side effect) so the motion enhancement is more pronounced at edges where the derivative of pixel/color change will be most extreme.

      The implications for CCTV/television camera based heartbeat pulse tracking are somewhat astounding/chilling when you think about it. It's essentially a video camera polygraph machine which makes you wonder if that's why nobody is talking about it. (You can also see the color change from blood flow to your face)

      http://hackaday.com/2012/06/04/tracking-small-changes-in-video-to-see-someones-pulse/

    4. Re:camera shake? by calzones · · Score: 1

      This is a mystery to me... color changes. The building changes color as it moves?
      We need a better TFA. The one linked explains pretty much nothing.
      Someone mod this guy up too.

      --
      Asking people to think is like asking them to buy you a new car
    5. Re:camera shake? by Whiternoise · · Score: 2

      No, the building stays the same colour. Very simply, consider a particular feature on the building. The location of that feature will shift between adjacent pixels in the image if the building moves relative to the camera. When this happens the pixels change colour (e.g. a 'sky' pixel might now be a 'building' pixel).

      The technique can be exploited for other things like blood flow, but in general things don't change colour as they move - unless they're travelling really fast

      And I've noticed this a lot on recent submissions, tons of second or third hand sources that aren't terribly useful.

      http://people.csail.mit.edu/mrub/vidmag/: Original source for Eulerian video magnification

    6. Re:camera shake? by Anonymous Coward · · Score: 0

      To be more precise.. there are subpixels in a bayer pattern. As the building moves, it'll occlude first a green subpixel, and then a red one, say. You can use this to get more resolution. And you can even go further and notice the subpixel STARTING to change in luminance and use that to get sub-subpixel superresolution. :)

    7. Re:camera shake? by Anonymous Coward · · Score: 1

      You're not being more precise. They're working with downsampled, post-bayer images. This isn't a cute hack around bayer patterns, it's hard math that's explained in the paper.

    8. Re:camera shake? by Neil+Boekend · · Score: 1

      It's similar to moire pattern analyses. Moire patterns have been used in position sensors for over a decade. In an old job I worked with a Heidehain moire pattern sensor that gave reliable contactless sub micrometer position data.

      --
      Well, I might have a way, but it only works on a semi spherical planet in a vacuum.
  2. just get a dslr by Anonymous Coward · · Score: 0

    you can see camera shake in images from taking pictures with a good enough dslr

    1. Re:just get a dslr by Anonymous Coward · · Score: 0

      We're trying to see building shake, not camera shake.

    2. Re:just get a dslr by Anonymous Coward · · Score: 0

      mount the camera on the building, have it photograph something with a semi long exposure, you will see the image blurred as the building is swayed by the *wind*

  3. Go to one of the higher floors in the Sears Tower by stox · · Score: 1

    On a windy day, the whole building sways. On the upper floors, you can really feel it. Last week it was so bad that we were getting sea sick from the motion.

    --
    "To those who are overly cautious, everything is impossible. "
  4. If you really wanted to see a building shake by Anonymous Coward · · Score: 0

    You should be in the vicinity when I'm ploughing my woman.

  5. Terrible soundtrack by lucm · · Score: 1

    I really wanted to look at the video and learn about this thing, but the soundtrack is so bad... I had to give up when it got to the "solo" that starts around 0:38. It sounds like they taped some random person who tried a few keys on a Casio keyboard on display at radio shack.

    I think the chilling screams of dogs being skinned live would have been less distracting.

    Someone should find the person who picked the soundtrack and pee on the windshield of their car.

    --
    lucm, indeed.
  6. John Hancock, Boston by turning+in+circles · · Score: 1

    Yes, you read that correctly in the summary, there actually is a John Hancock tower in Boston (240 m tall). It is shorter than the better-known one in Chicago (321 m tall, twice as many hits by Google). The Boston one looks very skinny, so maybe it shakes enough for the video gizmo stuff to pick up vibrations. (What me, make fun of MIT?)

    --
    Might as well face it I'm addicted to data.
    1. Re:John Hancock, Boston by Anonymous Coward · · Score: 0

      Yes it does, From the top floor on a windy day, you can feel that building sway also, even though it is only 60 floors.

  7. Old by klui · · Score: 1

    Saw this several years ago. https://www.youtube.com/watch?...

    1. Re:Old by Anonymous Coward · · Score: 0

      Yes, that link is old. However, this is a significant improvement. You can see your old videos right next to the new ones. The improvement is impressive!