3D Pre-stack Data Research Articles

Deep-learning-based local wavefront attributes and their application to 3D prestack data enhancement

The work presents a novel workflow to accelerate the estimation of local wavefront attributes (LWAs) from massive 3D prestack seismic data using deep learning (DL) focusing on data enhancement. A standard estimation method based on a semblance-based brute-force optimization provides good results but is time consuming. A modification of the U-net convolutional neural network, commonly used in image processing applications, is proposed to link the seismic data with the wavefront attributes. Color pixel image input for the neural network is generated through a straightforward seismic data regularization based on supergrouping followed by red, green, and blue encoding. The proposed workflow can be adapted to any 3D prestack seismic volume. Conventional semblance-based attributes estimation is required for the training step but only for approximately 1% of the total data. The prediction step is very efficient and reduces the overall run time significantly. The verification of the proposed approach is performed on challenging real land and marine data sets. As a result, DL-based estimation of LWAs accelerates computation up to 200 times compared to the standard method. The attributes from the proposed DL-based approach indicate an acceptable match compared with the brute-force semblance-based optimization results. Conventional and proposed estimation methods result in comparable prestack data enhancement results for more reliable seismic processing in challenging areas.

Prestack seismic random noise attenuation using the wavelet-inspired invertible network with atrous convolutions spatial pyramid

Convolutional Neural Network (CNN) is widely used in seismic data denoising due to its simplicity and effectiveness. However, traditional seismic denoising methods based on CNN ignore multi-scale features of seismic data in the wavelet domain. The lack of these features will decrease the accuracy of denoising results. To address this barrier, a seismic denoise method based on the wavelet-inspired invertible network with atrous convolutions spatial pyramid (WINNet_ACSP) is proposed. WINNet_ACSP follows the principle of lifting wavelet transform. The proposed method utilizes the redundant orthogonal wavelet transform to obtain frequency multi-scale information from noisy seismic data. Then predict update network (PUNet) extracts spatial multi-scale features of approximate and detailed parts. The sparse driven network (SDN) learns the complex multi-scale information and obtains sparse features. These sparse features are processed to eliminate random noise. Compared to standard convolution, the atrous convolutions spatial pyramid (ACSP) can extract more features. The redundant features are the key to ensure the precision of multi-scale information. Therefore, the introduction of ACSP in PUNet can guarantee the denoising effect of the network. WINNet_ASCP combines the characteristics of wavelet transform and neural network and has a high generalization. Besides, transfer learning is used to overcome the difficulty caused by the training sample size of seismic data. The training process includes pre-training and post-training. The former is trained to obtain the initial denoising network by natural image samples. The latter is trained with a small sample of seismic data to enhance stratigraphic continuity. Finally, the proposed method is tested with synthetic and field data. The experimental results show that the proposed method can effectively remove random noise and reduce the loss of detailed information in prestack seismic data. In the future, we will make further improvements on this basis and conduct experiments on 3D prestack data.

Evaluating strategies for estimation of local kinematic parameters in noisy land data: quality versus performance trade-offs

Abstract Modern land seismic data are typically acquired using high spatial trace density with small source and receiver arrays or point sources and sensors. These datasets are challenging to process due to their massive size and relatively low signal-to-noise ratio caused by scattered near-surface noise. Therefore, prestack data enhancement becomes a critical step in the processing flow. Nonlinear beamforming had proved very powerful for 3D land data. However, it requires computationally intensive estimations of local coherency on dense spatial/temporal grids in 3D prestack data cubes. We present an analysis of various estimation methods focusing on a trade-off between computational efficiency and enhanced data quality. We demonstrate that the popular sequential «2 + 2 + 1» scheme is highly efficient but may lead to unreliable estimation and poor enhancement for data with a low signal-to-noise ratio. We propose an alternative algorithm called «dip + curvatures» that remains stable for such challenging data. We supplement the new strategy with an additional interpolation procedure in spatial and time dimensions to reduce the computational cost. We demonstrate that the «dip + curvatures» strategy coupled with an interpolation scheme approaches the «2 + 2 + 1» method's efficiency while it significantly outperforms it in enhanced data quality. We conclude that the new algorithm strikes a practical trade-off between the performance of the algorithm and the quality of the enhanced data. These conclusions are supported by synthetic and real 3D land seismic data from challenging desert environments with complex near surface.

The Use of 3D Convolutional Autoencoder in Fault and Fracture Network Characterization

Conventional pattern recognition methods directly use 1D poststack data or 2D prestack data for the statistical pattern recognition of fault and fracture network, thereby ignoring the spatial structure information in 3D seismic data. As a result, the generated fault and fracture network is not distinguishable and has poor continuity. In this paper, a fault and fracture network characterization method based on 3D convolutional autoencoder is proposed. First, in the autoencoder training frame, 3D prestack data are used as input, and the 3D convolution operation is used to mine the spatial structure information to the maximum and gradually reduce the spatial dimension of the input. Then, the residual network is used to recover the input’s details and the corresponding spatial dimension. Lastly, the hidden features extracted by the encoders are recognized via k -means, SOM, and two-step clustering analysis. The validity of the method is verified by testing the seismic simulation data and applying real seismic data. The 3D convolution can directly process the seismic data and maximize the prestack texture attributes and spatial structure information provided by 3D seismic data without dimensionality reduction and other preprocessing operations. The interleaving convolution layer and residual block overcome low learning and accuracy rates due to the deepening of networks.

Interpolated multichannel singular spectrum analysis: A reconstruction method that honors true trace coordinates

The multichannel singular spectrum analysis (MSSA) reconstruction algorithm denoises and reconstructs seismic traces on a regular grid. We have developed a modified version of MSSA that can cope with denoising and reconstruction of traces with irregular coordinates. The proposed method, interpolated multichannel singular spectrum analysis (I-MSSA), connects off-the-grid observations to the desired gridded data via a linear interpolation operator. The algorithm consists of two steps. In the first step, we use the steepest-descent method to estimate the gridded data that honors off-the-grid observations. The second step guarantees convergence to a solution by applying the MSSA filter to the gridded data. The final solution is the reconstructed volume that honors off-the-grid observations. We use the algorithm to process synthetic and field data. We also provide an application in which 3D prestack data corresponding to an orthogonal survey are fully reconstructed using cross-spread gathers. We use I-MSSA to restore each subset individually. The output is a complete seismic volume described in a regular common-midpoint grid.

3D Kirchhoff depth migration algorithm: A new scalable approach for parallelization on multicore CPU based cluster

In this article, a new scalable 3D Kirchhoff depth migration algorithm is presented on state of the art multicore CPU based cluster. Parallelization of 3D Kirchhoff depth migration is challenging due to its high demand of compute time, memory, storage and I/O along with the need of their effective management. The most resource intensive modules of the algorithm are traveltime calculations and migration summation which exhibit an inherent trade off between compute time and other resources. The parallelization strategy of the algorithm largely depends on the storage of calculated traveltimes and its feeding mechanism to the migration process. The presented work is an extension of our previous work, wherein a 3D Kirchhoff depth migration application for multicore CPU based parallel system had been developed. Recently, we have worked on improving parallel performance of this application by re-designing the parallelization approach. The new algorithm is capable to efficiently migrate both prestack and poststack 3D data. It exhibits flexibility for migrating large number of traces within the available node memory and with minimal requirement of storage, I/O and inter-node communication. The resultant application is tested using 3D Overthrust data on PARAM Yuva II, which is a Xeon E5-2670 based multicore CPU cluster with 16 cores/node and 64GB shared memory. Parallel performance of the algorithm is studied using different numerical experiments and the scalability results show striking improvement over its previous version. An impressive 49.05X speedup with 76.64% efficiency is achieved for 3D prestack data and 32.00X speedup with 50.00% efficiency for 3D poststack data, using 64 nodes. The results also demonstrate the effectiveness and robustness of the improved algorithm with high scalability and efficiency on a multicore CPU cluster.

Seismic imaging by 3D partial CDS method in complex media

Abstract Seismic imaging in complex geological structures is state of art to deliver a high quality section for further geological interpretation. Conventional processing and imaging methods will fail in case of facing complex media, strong heterogeneity and poor quality data. Thus, conventional seismic imaging methods have to be improved inevitably or new approaches have to be extended to produce valid images. The partial common diffraction surface method is based on improving the common diffraction surface operator to handle lateral heterogeneities in complex media. In this study, the partial diffraction surface method was improved to 3D. This method enhances quality of the 3D pre-stack data to be used for further depth imaging. It can also resolve some of the possible ambiguities in geological interpretation from seismic images. This also will enhance the quality of the seismic images, where it suffers from conflicting dips problem. A time variant linear function defines the offset range for each zero offset sample in the stacking operator. This function controls the offset increment for the partial form of the operator in the time-midpoint-offset domain. It should be noted that the velocity model can be considered as smooth and simple as possible in this method for further depth imaging. The introduced method was applied on a 3D synthetic and a real land data set. Results from both data example show the ability of the new method in enhancing the quality of the seismic section in the presence of faults and lateral velocity heterogeneity. This enhancement achieved by a simple velocity model in a strong complex media. Consequently, depth migrated result of the real land data underwent a time conversion to create a time underground contour map, for comparison with the maps obtained with conventional methods. More details of the target formation were observed on the map used result of the proposed method for mapping.