Fault detection is a crucial step in seismic interpretation, which can be regarded as a segmentation task in computer vision. Existing deep learning methods train models using synthetic data. However, due to differences between synthetic and field data in signal-to-noise ratio (SNR), seismic resolution, and fault orientation, models trained on synthetic data may yield unreliable results when applied to field data. In this article, we assume that the features required for fault detection are irrelevant to nonfault features such as SNR and propose a cross-disentangled representation learning method for 3-D fault detection, called FaultCDR. FaultCDR comprises a fault encoder, a nonfault encoder, a seismic reconstructor, and a segmenter. It employs a cross-disentangled representation mechanism to decouple fault features and nonfault features. The cross-disentangled representation mechanism is achieved through the seismic reconstruction task of remixed fault/nonfault features and a self-supervised feature consistency task. The proposed orthogonal loss is used to ensure that fault features and nonfault features are irrelated. The decoupled pure fault features are finally fed into the segmenter for fault detection. Through intro-database and cross-database testing, we demonstrated the stability and generalization of FaultCDR in fault detection across different datasets. Comparative experiments with existing state-of-the-art (SOTA) fault detection methods reveal that FaultCDR achieves superior performance in both detection accuracy and visual quality.

Current data-driven salt body interpretation methods are mainly based on 2-D seismic slices and complete labeling training. The 2-D salt body prediction results of this kind of method lose the spatial continuity of salt body distribution after being restored to 3-D seismic space. With the difficulty in acquiring salt body labels in the field, it becomes crucial to use sparse 2-D labeling to guide the learning of 3-D networks. We have proposed a 3-D salt body segmentation method based on multiview collaborative regularization, called 3-D multiview co-regularization (SALT-MVCR). Innovatively, we designed a dual-view collaborative training paradigm for voxel-level seismic data and proposed a regional loss function applicable to 2-D sparse-salt body labeling, which solved the difficult problem of asymmetrically supervised sample learning. In addition, a cross-view prediction consistency loss was designed to improve the segmentation model’s understanding of the salt body information by restricting the parameter search space of a single view and solving the artifacts of the prediction result splicing problem. Experimental results show that after supervised training with only 1.56% of salt body labels, a Dice index of 90.6% has been achieved. The visualization of the 3-D salt body distribution also demonstrates that 3-D SALT-MVCR is capable of interpreting the complete salt body from the 3-D seismic body end-to-end and outperforms previous state-of-the-art methods in terms of segmentation performance.

The interpolation and reconstruction of missing traces are crucial steps in seismic data processing; moreover, it is also a highly ill-posed problem, especially for complex cases such as high-ratio random discrete missing, continuous missing, and missing in fault-rich or salt body surveys. These complex cases are rarely mentioned in current works. To cope with complex missing cases, we propose multidimensional adversarial generative adversarial network (MDA GAN), a novel 3-D GAN framework. It keeps the anisotropy and spatial continuity of the data after 3-D complex missing reconstruction using three discriminators. The feature splicing module is designed and embedded in the generator to retain more information of the input data. The tanh cross entropy (TCE) loss is derived, which provides the generator with the optimal reconstruction gradient to make the generated data smoother and continuous. We experimentally verified the effectiveness of the individual components of the study and then tested the method on multiple publicly available data. The method achieves reasonable reconstructions for up to 95% of random discrete missing and 100 traces of continuous missing. In fault and salt body enriched surveys, MDA GAN still yields promising results for complex cases. Experimentally, it has been demonstrated that our method achieves better performance than other methods in both simple and complex cases. Moreover, our network does not require training weights for each survey, the same weights it uses are applied to multiple surveys, significantly reducing time and computational costs, and we make the model publicly available onhttps://github.com/douyimin/MDA_GAN.

Data-driven fault detection has been regarded as a 3D image segmentation task. The models trained from synthetic data are difficult to generalize in some surveys. Recently, training 3D fault segmentation using sparse manual 2D slices is thought to yield promising results, but manual labeling has many false negative labels (abnormal annotations), which is detrimental to training and consequently to detection performance. Motivated to train 3D fault segmentation networks under sparse 2D labels while suppressing false negative labels, we analyze the training process gradient and propose the Mask Dice (MD) loss. Moreover, the fault is an edge feature, and current encoder-decoder architectures widely used for fault detection (e.g., U-shape network) are not conducive to edge representation. Consequently, Fault-Net is proposed, which is designed for the characteristics of faults, employs high-resolution propagation features, and embeds Multi-Scale Compression Fusion block to fuse multi-scale information, which allows the edge information to be fully preserved during propagation and fusion, thus enabling advanced performance via few computational resources. Experimental demonstrates that MD loss supports the inclusion of human experience in training and suppresses false negative labels therein, enabling baseline models to improve performance and generalize to more surveys. Fault-Net is capable to provide a more stable and reliable interpretation of faults, it uses extremely low computational resources and inference is significantly faster than other models. Our method indicates optimal performance in comparison with several mainstream methods.