The traditional fault time detection method is dependent on the sampling frequency and is easily affected by harmonics and attenuation oscillation. Thus a fault time detection method combining signal fitting and image comparison is proposed. First, an attenuation factor is introduced into the Fourier series fitting model. This can accurately fit the harmonics and attenuation oscillation components in the extended signal. Then, the differential Hash is used to compare the extended signal and the original fault signal in segments, and a median filter is used to remove the scattered pixels caused by the fitting error. Finally, the fault time is detected column by column. The simulation results show that the method can control the detection error at the pixel level, and can accurately detect the fault time at a low sampling frequency or when there are signals with harmonics and damped oscillations. The method is suitable for a variety of complex fault scenarios.