Sparse‐Bayesian‐learning‐based translational motion estimation of electromagnetic vortex imaging