Currently, there is an active interest in the introduction of neural networks (NN) in seismic data interpretation (for automatic correlation of horizons and faults, for forecasting reservoir properties, and for highlighting geological objects etc.). Highlighting geological objects is particularly interesting from the point of view of object-oriented interpretation. Standard interpretative approaches are largely subjective and require considerable time. Using of NN for outlining an object makes it possible to give interpretation process greater objectivity and prepare object for subsequent morphometric analysis.
The article discusses use of NN to identify paleochannels in process of interpreting seismic data and generating images of their conceptual geological models. Paleocannels are the best object for testing new approaches, since it is possible to project our knowledge about the features of modern river sedimentogenesis, morphology of rivers, patterns of their development in time and space on events captured in geological history, referring to method of "actualism" by Charles Lyell. The article analyzes two methodological approaches. The first one is to apply computer vision algorithms and a Holistically Nested Edge Detection algorithm based on a deep learning model using a convolutional NN. These algorithms have been tested on the example of color images of results of spectral decomposition and images obtained using eXchroma technology, on which paleochannels were identified during interpretation. The second approach is to use NN to generate an image based on a text description. Such popular networks as Midjourney, Problembo, Kandinsky have been tested. This approach will allow to generate images when searching for analogues of developed sedimentation models. The development of NN methods for selecting channels and generating images of possible sedimentation analogues seems to be an extremely promising direction.
References
1. Van der Walt S., Schonberger J.L., Nunez-Iglesias J. et al., scikit-image: image processing in Python, PeerJ, 2014, no. 2, DOI: https://doi.org/10.7717/peerj.453
2. Rogers M.S.J., Bithell M., Brooks S.M., Spencer T., VEdge_Detector: automated coastal vegetation edge detection using a convolutional neural network,
DOI: https://doi.org/10.1080/01431161.2021.1897185
3. Xie S., Tu Z., Holistically-nested edge detection, Int J Comput Vis, 2017, V. 125, pp. 3-18, DOI: https://doi.org/10.1007/s11263-017-1004-z
4. Alekseeva P.A., Kalugin A.A., Kir’yanova T.N., Identification of Tyumen formation paleochannels by using a neural network and 3D seismic data (In Russ.), Geofizika, 2022, no. 3, pp. 9-15.
5. Ol’neva T.V., Zhukovskaya E.A., Oreshkova M.Yu., Kuz’min D.A., Diagnostics of morphogenetic types of paleochannels on the basis of parameterization of seismic images (In Russ.), Geofizika, 2022, no. 2, pp. 17-25.
