How can solvePnPRansac be used with double values?

edit

I'm trying to use solvePnPRansac() to compute the 3D pose from a known 2D projection of the object. My values (intrinsic camera parameters, model points (3D), image points (2D)) are all double values. The normal solvePnP() does not complain about that.

However, when I try to use solvePnPRansac, it yields an error message:

Assertion failed < opoints.depth() == CV_32F >

How can I use this function with double values? (working with float does not give the desired results)