Multitask Seismic Inversion Based on Deformable Convolution and Generative Adversarial Network

Seismic inversion is crucial for oil and gas exploration. Recently, the development of deep learning (DL) has provided new means for the continuous improvement of seismic inversion. However, field seismic data exhibits nonstationarity and multiscale features due to wavelet attenuation and dominant frequency variations. Although the atrous spatial pyramid pooling (ASPP) module concatenated of dilated convolution greatly facilitates the multiscale feature learning ability of the network, the adjustment of the dilation rate, which is a hyperparameter, requires ample ablation experiments and is a tedious and time-consuming process. To alleviate these issues, the stacked multiple deformable convolutions (DConv) layers are employed as the feature extraction module to adaptively capture the multiscale correspondence between seismic data and elastic parameters in multitask seismic inversion. The sampling grids of DConv can automatically be modulated by adding a learnable offset. Thus, DConv can provide a flexible and effective receptive field, which is conducive to aggregating pivotal information of seismic data and improving the inversion accuracy. To further enhance the reliability and stability, the proposed method incorporates a multitrace to single-trace (M2S) strategy and the closed-loop Wasserstein generative adversarial network with gradient penalty (WGAN-GP) framework. Experiments show that the application of DConv to the inversion of P-wave velocity ( Vp ) and density ( rho ) yields superior transverse continuity and vertical resolution. Compared with ASPP, the MSE of the predicted profiles and the true models in the synthetic Marmousi2 experiment is degraded by 36% and 32%, respectively, and the MSE of the reconstructed seismic data and the real data is reduced by an order of magnitude for the field test.