3
$\begingroup$

sorry for my English, I write from Italy. I'll try to explain my problem. I'm tracking a footage with a solver error 0.2973. When I try to set the floor, the orientation of the camera is wrong (see screenshot). enter image description here So I tried to manually orient the camera, but even if well oriented respect to the xyz axes, geometry is incorrect as you can see from the screenshots below (the selected marker appears to be on the same plane but it is lower). enter image description here The video was recorded by a roof of a house so much higher than the road. It seems that Blender is not able to recognize this height and set the floor in a wrong point of view. Can someone help me solve this problem? Thank you

$\endgroup$
  • $\begingroup$ Can you include a google drive or similar link to your .blend file/video? $\endgroup$ – Shady Puck Jul 18 '16 at 11:46
  • $\begingroup$ Use one of the side views in orthogonal view (numpad 5) to reorient the scene. Read: blender.stackexchange.com/questions/10698/… $\endgroup$ – cegaton Jul 18 '16 at 14:47
  • $\begingroup$ Also related: blender.stackexchange.com/questions/28862/… $\endgroup$ – cegaton Jul 18 '16 at 14:57
  • $\begingroup$ Hi and thanks for answers. I oriented the camera just like that. The problem is that by rotating the camera (and all the markers), all track point that before the rotation were on the floor, after the rotation are located under the floor. Here is the link to download the project and the footage: 1drv.ms/u/s!AkNHGG_7aD9OlxzO_OZ0KU3scYf2 $\endgroup$ – Tyler Durden Jul 18 '16 at 16:42
4
$\begingroup$

The problem with your shot is that there is no parallax information. The camera is just panning, but there is no displacement. In other words the camera is just rotating but stationary in the same place. The normal camera solver cannot reconstruct the distance to the tracked points to do a proper 3d solution on this kind of shots.

enter image description here

You need to solve as tripod.

enter image description here

Note that when you do that there is no way to get a proper reconstruction of the space in 3D. The tracking points are then projected in a spherical way around the camera.

enter image description here

Tripod solve will not give you real 3D information.

From the wiki:

tripod can be used for footage where the camera does not move and only rotates. Such footage can't be tracked with a generic solver approach, and it's impossible to determine the actual feature points in space due to a lack of information. So this solver will solve only the relative camera rotation and then reproject the feature points into a sphere, with the same distance between feature and camera for all feature points.

To integrate 3D elements into the shot you're going to need to know the actual lens/sensor used and try to calculate the camera height manually.

$\endgroup$
  • $\begingroup$ Further reading: blender.stackexchange.com/questions/42329/… $\endgroup$ – cegaton Jul 18 '16 at 17:33
  • $\begingroup$ Thanks, I figured that the problem was due to no parallax information. Unfortunately I had also tried to solve as tripod, but the result was about the same. Thank you all for the replies. $\endgroup$ – Tyler Durden Jul 18 '16 at 17:39
  • $\begingroup$ Blender has no way of guessing if the points are closer or further away with no parallax. $\endgroup$ – cegaton Jul 18 '16 at 17:40
  • $\begingroup$ try using blam to determine the proper perspective. github.com/stuffmatic/blam $\endgroup$ – cegaton Jul 18 '16 at 17:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.