How about 2 (or more) separate cameramen with recording GPS's to create a synthetic aperture.
The cameras need to be looking the *exact* same direction with the *exact* same orientation. Think of looking straight forward with your eyes. Compare that with crossing your eyes. When you look straight forward, you get two slightly different variations on the same image that you can make sense of to create depth, because pieces of each image overlap. When you cross your eyes, it's just two very different images that you can't make sense out of.
The logic to writing a program that can create depth from two images is very close to what our brains do. You need to find objects in each image and then see how much they overlap horizontally. That amount of overlap gives your brain (or the program) the info it needs to calculate distance (relative to other objects in view). The images have to be very, very similar to perform depth calculations.
Unfortunately, cameras aren't as high resolution as our eyeballs and it's tough to write software as smart as our brains, so trying to accurately judge depths across wide spans or great distances is basically not an option with current technology.
How about 2 (or more) separate cameramen with recording GPS's to create a synthetic aperture.
The cameras need to be looking the *exact* same direction with the *exact* same orientation. .
No. Think tomography. The cameras need to be looking from different directions. The math is very well established (I have even written some iterative reconstruction software myself for analyzing X-ray data). All you need to know is where the cameras are at any instant (hence the need for the recording GPS)
This is getting way off track. If tomography was even applicable for this application, and if you did film from two angles with GPS on the flyers -- it'd be nearly impossible to calibrate time, GPS, and images in the video to result in useful data.
For exceptionally shaped formations, make an exception and do it by hand. For anything that has a pattern, we can feed it to a computer.
Baby steps. Start with formations on one level with basic patterns (where triangles can be formed between jumpers), and then let's move on from there.
(This post was edited by LetsGoOutside on Apr 21, 2010, 1:56 PM)
This is getting way off track. If tomography was even applicable for this application, and if you did film from two angles with GPS on the flyers -- it'd be nearly impossible to calibrate time, GPS, and images in the video to result in useful data.
For exceptionally shaped formations, make an exception and do it by hand. For anything that has a pattern, we can feed it to a computer.
Baby steps. Start with formations on one level with basic patterns (where triangles can be formed between jumpers), and then let's move on from there.
I was responding to your comment that you would not trust your method for 3-d formations.
Quote:
You can do some cool illusions with it, but it's not something I would trust to judge 3D formations.
A tomographic method would be just fine for 3-d formations. So it's more difficult
(This post was edited by kallend on Apr 21, 2010, 6:01 PM)
9waydiamond1.jpg
