The plugin tries to guess Position, Rotation and beamer intrinsics by assuming values for them.
these 9 values are used to project the input (3d points from the scene for example) and comparing the result to the mapped 2d points you provided (screen space). there will be some error, but the NelderMeadSolver takes care of reducing that error continuously (by changing the guessed values slightly) until the discrepancy is small enough and the guessed values can be taken for real.
On a side note: There are some issues related to convergence to a local minimum, thats why there is all that patch stuff to the right to detect if we are maybe dealing with a local minimum.
great!
nice! does it work different than cv´s camera calib? looks quite familiar but a simpler. i guess the equations are kind of the same?
very cool! based on image pack?
No, it only depends on the MS Solver Foundation that ships with .Net anyway.
https://github.com/velcrome/vvvv-Dome/blob/master/modules/Mapping/Solver%20(3d)/plugins/ValueSolver/ValueSolverNode.cs
The plugin tries to guess Position, Rotation and beamer intrinsics by assuming values for them.
these 9 values are used to project the input (3d points from the scene for example) and comparing the result to the mapped 2d points you provided (screen space). there will be some error, but the NelderMeadSolver takes care of reducing that error continuously (by changing the guessed values slightly) until the discrepancy is small enough and the guessed values can be taken for real.
On a side note: There are some issues related to convergence to a local minimum, thats why there is all that patch stuff to the right to detect if we are maybe dealing with a local minimum.