Self-supervised delineation of geologic structures using orthogonal latent space projection

We have developed two machine-learning frameworks that could assist in automated lithostratigraphic interpretation of seismic volumes without any manual hand labeling from an experienced seismic interpreter. The first framework is an unsupervised hierarchical clustering model used to divide seismic images from a volume into a certain number of clusters determined by the algorithm. The clustering framework uses a combination of density and hierarchical techniques to determine the size and homogeneity of the clusters. The second framework consists of a self-supervised deep learning framework to label regions of geologic interest in seismic images. It projects the latent space of an encoder-decoder architecture onto two orthogonal subspaces, from which it learns to delineate regions interest in the seismic images. To demonstrate an application of both frameworks, a seismic volume is clustered into various contiguous clusters, from which four clusters are selected based on distinct seismic patterns: horizons, faults, salt domes, and chaotic structures. Images from the selected clusters are used to train the encoder-decoder network. The output of the encoder-decoder network is a probability map of the possibility that an amplitude reflection event belongs to an interesting geologic structure. The structures are delineated using the probability map. The delineated images are further used to posttrain a segmentation model to extend our results to full vertical sections. The results on vertical sections indicate that we can factorize a seismic volume into its corresponding structural components. Finally, we find that our deep learning framework can be modeled as an attribute extractor and we compare our attribute result with various existing attributes in the literature and determine the competitive performance with them.