I'm looking for some help trying to solve the geometry / camera positioning for a show clip. After many days of testing and tweaking I still can't get a matching result and I just don't understand why. The tracking itself seems pretty stable, but the actual location and orientation of the camera does not match at all. Target is to place a pretty big bed (width: ~1m, length: ~2m, height: ~2.5m) into the room.

In addition to some other randomly distributed markers, there are 4 markers on the ground, which mark exactly the middle of the outer posts of the bed. The distance from left to right is exactly 1980mm and the distance from front to back is exactly 974mm.

Footage / Tracking

Technical Information:

  • Camera: Sony Alpha 7 III (35.6mm, 3:2 ratio, Full Frame Sensor)
  • Lens: 28mm (Sony SEL28F20)
  • Footage: 4K (16:9, 3840x2160), 25 FPS (PAL)

I tried:

  • Using other lenses: 24mm (Sony SEL24F14GM), 50mm (Sony SEL50F18F)
  • Manually match camera position and orientation
  • Using existing geometry as additional markers
  • Using automatically generated markers ("detect features")
  • Using less markers
  • Let the solver automatically calculate the focal length, K1 & K2
  • Solving orientation via "Wall" - the result didn't make sense at all
  • Solving orientation via "Floor" - the result was at least somewhere around where it should be
  • The functions "Set X-Axis" and "Set Y-Axis" result in swapping the matching orientations by invalidating the respective other axis
  • Manually played around with the lens distortion values (K1 & K2)

The by far best result I got was by manually placing the camera, but after around 5-6 hours of incrementally adjusting its position and orientation, I still didn't get a 100% match: 3D Space

I uploaded my current blender project file and the associated video footage via Dropbox, if someone may want to have a look on his/her own.

Sometimes, after trying to solve the problem, blender got completely confused (just like me) and didn't allow to use any tracks to set the orientation ("Three tracks with bundles are needed to orient the floor"; solving the camera motion over and over again didn't fix it, so I threw away the whole solution and started over).

I would be very pleased about support in solving the problem.

Thank you.


1 Answer 1


Just moving the camera around, panning and tilting while remaining on the same spot will not work, so don't waste any more time trying to match anything.

You will not be able to get 3d information from a stationary camera.

If there is no displacement (as in: taking a few steps while filming the object) there is no way for blender to determine 3d from parallax. To estimate depth, it is necessary to see the objects from different places.

If the camera is static you can use "Tripod" solve (https://docs.blender.org/manual/en/latest/movie_clip/tracking/clip/editing/solve.html#tripod), but note that the tracked elements will not yield any 3D information, only camera rotation at best. The tracked elements will only be projected in a sphere around the camera.

For detailed explanation read through these related posts:

Problem with camera orientation in motion tracking

How can I get better results when doing camera motion tracking?

Motion tracking is oriented incorrectly; no depth

Using the same tracking marks make another shot (or series of photographs) and walk around that corner of the room. Then you will be able to make a proper camera solution and have 3D information.


Not the answer you're looking for? Browse other questions tagged .