In the inverse synthetic aperture radar (ISAR) imaging, the non-uniform rotation of the targets can bring phase errors relevant to scatters’ positions. The phase errors can not be expressed by one uniform function. But the scatters’ phase precision is very important to phase emendation of ISAR’s autofocus. For the accelerating targets, an autofocus algorithm based on fractional Fourier transform (FRFT) is proposed. FRCFT is applied to process the signals in the selected range unit, and extract the scatter’s signal with the maximum energy. this selected scatter signal is taken as the reference signal to eliminate the plane motion phase errors. By the same method, the scatters’ signals with maximum energy are also extracted on several selected range units. And modulated frequency signals are weighted on the average to get the optimized estimated values of the non-uniform rotation parameters. Based on the parameters, the signals are sampled non-uniformly to eliminate the phase errors. Finally, the simulation results demonstrate that the algorithm can eliminate the phase errors very well.