Full waveform inversion (FWI) is one of the most accurate velocity inversion tools.It can be used to obtain high-precision subsurface structures through iterative inversion and provide a more accurate velocity field for prestack imaging technology,thus satisfying the increasingly complex demand for petroleum exploration and development.However,FWI requires accurately estimated source wavelets,which are very difficult to extract from the seismic data collected in the field.Furthermore,in the inversion process,there is a strong nonlinear relationship between model parameters and observed data,which is liable to induce cycle skipping.To overcome the cycle skipping and the difficulty in extracting seismic wavelets in the inversion process of medium-low wavenumber,this paper developed a multi-scale waveform inversion method using combined source encoding.Specifically,the wavelets and seismic data were combined by applying time-shift stacking,and gradients can be determined through only one calculation of reverse time migration.According to the calculation results using a model,the FWI method using combined source encoding can be used to achieve multi-scale inversion and obtain more stable inversion results.Meanwhile,relatively accurate inversion results can be obtained by combining this method with a source-independent method.