I do know there are some motion controllers that offer a closed loop feedback system. Vital System's HiCON can for one.
Having never used a 3D vision software before, I have no idea what the output data would look like, but if it is accurate enough to distinguish X, Y, and Z positons, you may be able to convert those into data that the feedback system can use. The HiCON uses encoder feedback, whether that is from an rotary encoder or from a linear encoder, it is still A, A-, B, B- commands. If your software could be converted to steps to feedback into this you may be able to get something to work. Just a thought... throwing ideas out there. Maybe one will stick.