Unusable Data

[FlantasticDan]

The data coming out of the mocapSolver.py does not result in data which is usable. Something must be wrong with the math.

Ideas:

• Work backwards with pre-tracked 3D data from the calibration stages of camera alignment.
• Check radians and angle of view operations

[FlantasticDan]

BREAKING NEWS

• Blender orients the world as Z-up for no particularly good reason 😡

• The code had been written assuming the data was presented in the normal "industry standard" Y-up

[FlantasticDan]

Update

• Updating to the Z-up axis improved accuracy to a degree but...

• The camera values are reported in world space which is not equivalent to local space values used to calculate the Angle of View adjustment.

[FlantasticDan]

Mathutils from Blender 2.78

• Replacing the original rotation calculator with a new iteration based on the mathutils module has solved the local vs. world space issue.
  • Installing the module requires a community fork off of the Blender source: https://github.com/majimboo/py-mathutils

Going Forward

• Correct rotation data still does not result in the correct placement of a second point to define the line due to a lack of orientation information

[FlantasticDan]

Solved!

• The solution was reached by projecting an imaginary point 1 unit in front of the camera at its spawn point, rotating that about the x, followed by y, followed by z axis according to the calculated Angle of View calibration. The rotated point is then translated to the current camera position to form a line defined by this projected point and the camera's position.

  • The mathematics for this operation come courtesy of a write up by Paul Bourke: http://paulbourke.net/geometry/rotate/

    • The code implementation is a modified version of a script written by Bruce Vaughan originally for use in SDS/2 design software.