| stereo triangulation in IDL [message #81417] |
Sat, 15 September 2012 10:31  |
derkleinepilz
Messages: 3
Registered: September 2012
|
Junior Member |
|
|
Hey,
I want to calculate the top height of a cloud using pictures of two satellites, but I'm kinda new to this stuff and figured that stereo triangulation should do the trick. Does anyone know if there already is a procedure doing stereo triangulation out there? That would be great, 'cause it looks really complicated. I wouldn't know how to programm something like this.
regards, Haiko
|
|
|
|
|
|
| Re: stereo triangulation in IDL [message #81494 is a reply to message #81417] |
Mon, 17 September 2012 23:53  |
derkleinepilz
Messages: 3
Registered: September 2012
|
Junior Member |
|
|
Am Montag, 17. September 2012 23:52:03 UTC+2 schrieb Klemen:
> On Thursday I will submit the discussion paper for ACP, then I can post the link here. But a shorter version is: everything is based on parallax. :)
>
>
>
> I have both datasets in the same projection plane. I use moving window analysis that correlates both datasets. The pixel pair with the highest correlation is the one I am interested. Shifts are then just the differences i the array indices. For each pixel in the array I know the lon, lat, I know the satellite position, so I can generate two lines. From the intersection point between these lines I can computethe height.
>
>
>
> Processing over pyramids is faster and it considers the larger scale features and the small scale features. Correlation is first analysed on the coarse level data. Once I have the indices I can just proceede to more detailed pyramid, where the previously determined shifts are the basis for the mowing window analysis.
>
>
>
> So much in short, if you need more details, I can provide them in the following days!
>
> Klemen
That would be great. My plan was to use epipolar lines in my stereo matching process, because it would be faster than cross-correlating the whole datasets. But it seems like I need a lot of camera parameters for this. Like you I have the lons and lats for each pixel and the satellite position. And that is where I'm stuck. I just don't know how to compute the height with that intersection point. Maybe it's just trivial (because on all the papers I've read about that topic, it's all about that correspondence point problem not about the heightcalculations- most of them just end with "and now we can compute the height". So I decided to do it like this: http://solarphysics.livingreviews.org/open?pubNo=lrsp-2011-5 &page=articlesu6.html
Well anyways, I'm looking forward to your discussion paper.
|
|
|
|
|
|
| Re: stereo triangulation in IDL [message #88682 is a reply to message #81417] |
Sun, 01 June 2014 12:32 |
George[2]
Messages: 1
Registered: June 2014
|
Junior Member |
|
|
|
Hi! Its interesting. Depth/height can be calculated via measuring the parallax in two (or more) overlapping images of the same object. This will give you the result, if the the measured object does not move in time (because the two images taken in different time. Clouds are moving in fact. The question is how much noticable is this movement compared to the satellite altitude, and also as clouds are not solid objects their extension in object space is not strict, you cannot model this. Cheers.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
