Abstract:The seismic multi-parameter full waveform inversion combining wave velocity and medium density can fully utilize the kinematic and dynamic information of seismic data to reconstruct the parameters of underground medium, which plays an important role in geological interpretation of seismic data. However, the multi-parameter full waveform inversion of seismic wave is highly dependent on the initial model, and there is crosstalk noise between different parameters, which restricts the accuracy of the inversion results. Therefore, this paper proposes a multi-parameter multiplier inversion framework based on the multiscale strategy. Firstly, low-frequency inversion was carried out to obtain large-scale information in the model space; subsequently, Lagrangian multipliers were introduced to improve the inversion objective function, and the iterative update was completed; finally, the low-frequency inversion results were used as the initial high-frequency model for band by band inversion. The experiments of the simple transmission model and the 1994BP standard model show that the multiplier inversion algorithm based on multiscale strategy has good robustness and applicability. Even if the initial model is poor, it can still obtain better inversion results, and improve the inversion accuracy of density without reducing the velocity inversion accuracy.