www.ssontech.com

[borgus]

Hi guys!
i've searched the internet for shots to track and found cool plates at:
http://www.hollywoodcamerawork.us/trackingplates.html

i understand the benifit of survey shots, but what is the
extraction-dots shot for? to determine lens-distortion??

seb.

[borgus]

no one knows what thats for? can't believe that....

[GraphicsKid]

What are you talking about? Where are these extraction dots?

EDIT: Look at the soda can tracking one. It's used to determine the focal length very precisely (ie, exact measurements know from dots work much better than the actual thing you're tracking).

[borgus]

ahhh, now i got it! thx!

than i can determine the focal length of the main camera and can solve
the positions of the widness cameras, right? unfortunately there's no
real calibration for the tree cameras so i have to "indirectly" solve the other
cameras. i've tried that but the results are not that good.

is there an other way to determine the position of the three cams? one-frame
alignment isn't accurately enough.

sebastian

[GraphicsKid]

Woa woa woa you lost me... 3 cameras? What is it you're trying to do? How is your shot set up?

[borgus]

Hi GraphicsKid!
I got three cameras (main cam, widness cam 1 & 2). theres no calibration
shot except of the "dots extraction" shot.
my approach was to track some almost static points at the forehead to solve the main cam as a moving camera.

this is working very fine (the points in 3d space are looking very good and the error rate is far below .9). my approch was to add the other two shots and track the same points "indirectly" with connections to the main shot.
this is, from a error-rate point of view, fine.

then i run the calibration script, kill the path etc (the same way as it is descripted in the manual). then i turn the solving off (for the three cameras) and add moving objects fore each cam. i solve the points as 'individual mocap" and get results (but its not accurate enough).

i'm just asking myself how the guys at "hollywood camera work" solve the initial camera positions of the three cameras without having a calibration-shot for each cam....

[GraphicsKid]

Oh witness camera! I kept thinking you were trying to say wideness camera.

Unfortunately I've never worked on any major projects, usually me just running around in my backyard with a somewhat decent camcorder, so I have never been able to try a multiple camera setup like that.

If the solve is more accurate without the extra cameras, why are you incorporating them into the shot?

Or perhaps they are using the same type of camera for each shot and they only need to do the focal length check to one of the cameras. I'm just guessing.