Converted a number of objects in the frame order target functions to numpy.float16 types.
These were originally float64. These objects relate to the 3D structure which is not very accurate
anyway. This should help with the movement of data for the multi-processor optimisations.
