Prev: Problem with Matlab on Ubuntu Linux
Next: Rearranging arrays.
From: Laith on 27 May 2010 13:39
Dears,
Kindly Can you help me in Calculate the real distance between an object and a camera( that we are seeing the object through it, for any object in any snapshot. in using matlab & a camera (320*240) Resolution.
Can you give me an example please?
Thanks in advance.

Reagard,
Laith
From: David Young on 27 May 2010 14:00
Actually this is an unsolved problem. Indeed, there is no general solution, though in some cases it is possible to make estimates - for example if the scene includes something of known size, shape and orientation that can be used for calibration, and if the other elements of the scene are in some kind of regular arrangement with respect to it. Some inferences can be attempted from a single image, using, for example, linear perspective and texture gradients, but most practical solutions use multiple images from different viewpoints (stereo or motion).

For several decades, attempts to estimate the third dimension from one or more 2D images were seen as the core goal of computer vision. Now, the field has moved on, but this is still an area of vigorous research.
From: ImageAnalyst on 27 May 2010 14:44
You can't tell. You have an image of a galaxy. How can you tell if
it's an actual galaxy millions of light years away that you're seeing
from your camera hooked up to your telescope, or if it's just a photo
of a galaxy taped to your wall that's only a few feet away? You
can't. Like David said, you can make estimates if you make some
assumptions. So now how do you want to proceed? What do you really
need to accomplish? If you really need the distance, why don't you
try something like a rangefinder device ("tape measure" -like tool)
like you'd buy at the hardware store?
From: Bruno Luong on 27 May 2010 15:03
ImageAnalyst <imageanalyst(a)mailinator.com> wrote in message <90cc8db3-bf74-4518-ac6c-977ee7f3ae04(a)d12g2000vbr.googlegroups.com>...
> You can't tell. You have an image of a galaxy. How can you tell if
> it's an actual galaxy millions of light years away that you're seeing
> from your camera hooked up to your telescope, or if it's just a photo
> of a galaxy taped to your wall that's only a few feet away? You
> can't.

Well in principle you actually could, it's just difficult. A camera has only one plane called focus plane where the object is sharp. So in principle the "blurring" of the object image can tell where it's located from the focus plane (you also need to know the aperture when the photo is taken).

IA' example of stars is tricky because their light comes "almost" collimated (plane wave front), but they must still have tiny different of curvatures depending on they distance. But in the ideal world where Gaussian noise are always 0, we could.

Bruno
 | 
Pages: 1
Prev: Problem with Matlab on Ubuntu Linux
Next: Rearranging arrays.