GNGTS 2021 - Atti del 39° Convegno Nazionale

GNGTS 2021 S essione 3.1 400 Conclusions In this work, we presented a preliminary study of the MultiResolution U-Net as deep prior for interpolating seismic data to place both sources and receivers on the same regular dense grid. The network is able to extract significant features from time slices at different scales, but the interpolation results may be prone to aliasing. To overcome this limitation, we modify the network architecture by implementing 3D convolution kernels and processing adjacent time slices. This extension makes the CNN able to exploit the self-similarity of adjacent time slices, boosting the reconstruction performance. Moreover, we exploit the source-receiver reciprocity of the acquisition geometry by modifying the network output in order to force the reconstructed time slices to be symmetric. This physics- driven constraint brings an additional performance gain. Through numerical simulations we have proven the effectiveness of our methodology by reconstructing common receiver gathers with shots placed every 20m starting from common receiver gathers with shots placed every 80m. Future work will be devoted to further analyzing this interpolation ratio and tackling 3D geometries References Gülünay, N., 2003, Seismic trace interpolation in the fourier transform domain: Geophysics, 68, 355–369. Kong, F., V. Lipari, P. Bestagini, and S. Tubaro, 2020, A deep prior convolutional autoencoder for seismic data interpolation: Presented at the 82nd EAGE Conference and Exhibition 2020. Liu, Q., L. Fu, and M. Zhang, 2019, Deep-seismic-prior-based reconstruction of seismic data using convolutional neural networks, Geophysics, 86, 2, 2021. Mandelli, S., F. Borra, V. Lipari, P. Bestagini, A. Sarti, and S. Tubaro, 2018, Seismic data interpolation through convolutional autoencoder, in SEG Technical Program Expanded Abstracts 2018: Society of Exploration Geophysicists, 4101–4105. Oliveira, D., R. Ferreira, R. Silva, and E. Brazil, 2018, Interpolating seismic data with conditional generative adversarial networks: IEEE Geoscience and Remote Sensing Letters, 15, 1952–1956. Pham, N., and S. Fomel, 2019, Seismic data interpolation using cyclegan: Presented at the SEG International Exposition and Annual Meeting, Society of Exploration Geophysicists. Ronneberger, O., P. Fischer, and T. Brox, 2015, U-net: Convolutional networks for biomedical image segmentation: International Conference on Medical image computing and computer-assisted intervention, Springer, 234–241. Spitz, S., 1991, Seismic trace interpolation in the fx domain: Geophysics, 56, 785–794. Ulyanov, D., A. Vedaldi, and V. Lempitsky, 2018, Deep image prior: Presented at the The IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Wang, B., N. Zhang, W. Lu, and J. Wang, 2019, Deep-learning-based seismic data interpolation: A preliminary result: Geophysics, 84, V11–V20.