MLPnP solver modified for accurate scale recovery at large ranges
This is a modified implementation of the original Maximum Likelihood Perspective-n-Points solver. For details on the orginal algorithm, please refer to MLPnP - A Real-Time Maximum Likelihood Solution to the Perspective-n-Point Problem
##Modification: Scale recovery for linear system solution is done with the intermediate singular value rather than a coarse three column norm that breaks down at larger relative ranges.
To use the modified MLPnP, use MLPnP_mod(points3D, v, cov)
For original MLPnP, use MLPnP(points3D, v, cov)
... Description for modification, and Monte-Carlo analysis coming soon...
If it is after November 2020 and I still haven't updated, please raise an issue and I will quickly update everything. I'm just in normal circumstances, lazy :)