Camera matching

   6820   2   2
User Avatar
Member
319 posts
Joined:
Offline
Hi guys,

I've tried to search the web and odforce but I can't find an article/tutorial on camera matching in Houdini, at least I can't find one which I understand. I want to do a camera projection of footgae on to a CG building. When shooting my footage I took the following data:-

1. Camera - Canon XL2
2. Lens - 50mm
3. Camera height
4. Camera tilt
5. Distance of camera from building

With this data I thought it would be fairly simple to match my CG camera to my real world data, but alas, no. I specifiy 50mm as the focal length, the tilt in the transform info of the camera, the resolution to that of my footage (720 x 576) and all other details that I took and the match is extremely poor. Can someone tell me what I am doing wrong?

Thanks.
User Avatar
Member
306 posts
Joined: July 2005
Offline
My guess is you'll probably need to eyeball the aperture until it matches. I don't know how much you can rely on DV cam's focal lengths as meaningful units anyway.

Cheers,
r.
“First things first – but not necessarily in that order”
– The Doctor
User Avatar
Member
387 posts
Joined: July 2005
Offline
Have you've read the Help in the Houdini Browser about Camera Matching? I thought that was fairly good for starters.
hdox/houdini/content/commands/guide/exp_matchcam.html

Note the last comment (in the link above):
fovx is not dependent on resx, resy, or aspect, but fovy is.
You can verify this for yourself by attaching a unit-spaced grid to the camera at the focal length distance and the number of units in the X direction will be exactly equal to the aperture. The number of grid units in the y direction ( apy) will be dependent on resx, resy and aspect.
Rangi wrote something about this a while back, which appears to be catalogued at the Barr-Price Institute's Dept of Ascii Communication: http://www.sidefx.com/index.php?option=com_mailarchive&Itemid=212&view=WEB&msgid=004c01c32620$32394990$0b01000a@dkp.com&perpage=20&revdate=off [sidefx.com]

Alternately, get your camera params roughly right, and then render a unit-spaced wireframe grid with them (Grid Sop + Wireframe Sop). Don't change the grid. Render it and then comp that grid image over your camera image. Note how close you are. adjust the camera parms & render again until it lines up. A bit clunky, but there's one way which avoids the maths. :roll:
''You're always doing this: reducing it to science. Why can't it be real?'' – Jackie Tyler
  • Quick Links