Wireless power transfer systems can be utilized to advance electromobility, especially in local public transport. These systems emanate low-frequency stray magnetic fields that expose living beings in the vicinity and induce potentially harmful electric fields into the exposed biological tissues. To ensure personal protection, ICNIRP has established basic restrictions on the body-internal electric field strength that should not be exceeded. Since this field strength cannot be measured, numerical schemes such as the Scalar-Potential Finite Difference scheme need to be utilized to determine the induced body-internal electric field distribution. This numerical scheme allows a near real-time body-internal electric field determination but requires as input a fine sampling of the magnetic flux distribution in the domain of potential human exposure. The works' objective is to provide a component for deriving the large number of required magnetic flux densities from free-space field measurements in near real-time, required for direct body-internal electric field determinations in real exposure assessment environments. By applying a highly parallelized approximation procedure using radial basis functions on NVIDIA GPUs, the required magnetic flux densities for different human voxel models with an ICNIRP-recommended edge length of 2 mm can be derived in a few milliseconds.