Seismic image registration using multiscale convolutional neural networks

Geophysics ◽  
2020 ◽  
Vol 85 (6) ◽  
pp. V425-V441
Author(s):  
Arnab Dhara ◽  
Claudio Bagaini
Keyword(s):  
Image Registration ◽  
Optical Flow ◽  
Synthetic Data ◽  
Time Lapse ◽  
Training Data ◽  
Supervised Machine Learning ◽  
Shift Function ◽  
Data Sets ◽  
Advantages And Disadvantages ◽  
Seismic Image

Aligning seismic images is important in many areas of seismic processing such as time-lapse studies, tomography, and registration of compressional and shear-wave images. This problem is especially difficult when the misalignment is large and varies rapidly and when the images are not shifted versions of each other because they are either contaminated by noise or have different phase or frequency content. In addition, the images may be related by multidimensional vector-valued shift functions. We have developed a fast, scalable, and end-to-end trainable convolutional neural network (CNN) for seismic image registration. The concept of optical flow is widely applied to the problem of image registration using variational methods. Recent developments in the field of computer vision have shown that optical flow estimation can be formulated as a supervised machine learning task and can be successfully solved using CNNs. We train our CNN, SeisFlowNet, on images warped with known shifts and corrupted with noise, frequency, and phase perturbations. We evaluate the promising performance of the trained SeisFlowNet with synthetic data sets where the shift function is known and the images are contaminated with noise and other perturbations. The accuracy of the results obtained with SeisFlowNet is favorably compared with two other popular methods for seismic registration: windowed crosscorrelation and dynamic image warping. Further, we highlight the principles adopted to create training data sets and the advantages and disadvantages of the method.

SaltSeg: Automatic 3D salt segmentation using a deep convolutional neural network

2019 ◽  
Vol 7 (3) ◽  
pp. SE113-SE122 ◽  
Author(s):  
Yunzhi Shi ◽  
Xinming Wu ◽  
Sergey Fomel
Keyword(s):  
Large Scale ◽  
Model Building ◽  
Ground Truth ◽  
Velocity Model ◽  
Training Data ◽  
Data Sets ◽  
Validation Data ◽  
Data Set ◽  
Seismic Image ◽  
Data Generator

Salt boundary interpretation is important for the understanding of salt tectonics and velocity model building for seismic migration. Conventional methods consist of computing salt attributes and extracting salt boundaries. We have formulated the problem as 3D image segmentation and evaluated an efficient approach based on deep convolutional neural networks (CNNs) with an encoder-decoder architecture. To train the model, we design a data generator that extracts randomly positioned subvolumes from large-scale 3D training data set followed by data augmentation, then feed a large number of subvolumes into the network while using salt/nonsalt binary labels generated by thresholding the velocity model as ground truth labels. We test the model on validation data sets and compare the blind test predictions with the ground truth. Our results indicate that our method is capable of automatically capturing subtle salt features from the 3D seismic image with less or no need for manual input. We further test the model on a field example to indicate the generalization of this deep CNN method across different data sets.

scholarly journals Time-lapse difference static correction using prestack crosscorrelations: 4D seismic image enhancement case from Ketzin

Geophysics ◽  
2014 ◽  
Vol 79 (6) ◽  
pp. B243-B252 ◽  
Author(s):  
Peter Bergmann ◽  
Artem Kashubin ◽  
Monika Ivandic ◽  
Stefan Lüth ◽  
Christopher Juhlin
Keyword(s):  
Seismic Data ◽  
Time Lapse ◽  
Data Sets ◽  
Storage Site ◽  
Data Set ◽  
Near Surface ◽  
Static Correction ◽  
Seismic Image ◽  
Static Corrections ◽  
4D Seismic

A method for static correction of time-lapse differences in reflection arrival times of time-lapse prestack seismic data is presented. These arrival-time differences are typically caused by changes in the near-surface velocities between the acquisitions and had a detrimental impact on time-lapse seismic imaging. Trace-to-trace time shifts of the data sets from different vintages are determined by crosscorrelations. The time shifts are decomposed in a surface-consistent manner, which yields static corrections that tie the repeat data to the baseline data. Hence, this approach implies that new refraction static corrections for the repeat data sets are unnecessary. The approach is demonstrated on a 4D seismic data set from the Ketzin [Formula: see text] pilot storage site, Germany, and is compared with the result of an initial processing that was based on separate refraction static corrections. It is shown that the time-lapse difference static correction approach reduces 4D noise more effectively than separate refraction static corrections and is significantly less labor intensive.

Discrimination between pressure and fluid saturation changes from marine multicomponent time‐lapse seismic data

Geophysics ◽  
2003 ◽  
Vol 68 (5) ◽  
pp. 1592-1599 ◽  
Author(s):  
Martin Landrø ◽  
Helene Hafslund Veire ◽  
Kenneth Duffaut ◽  
Nazih Najjar
Keyword(s):  
Pore Pressure ◽  
Seismic Data ◽  
Water Saturation ◽  
Synthetic Data ◽  
Time Lapse ◽  
Seismic Survey ◽  
Data Sets ◽  
Amplitude Variation With Offset ◽  
Fluid Saturation ◽  
Pressure Changes

Explicit expressions for computation of saturation and pressure‐related changes from marine multicomponent time‐lapse seismic data are presented. Necessary input is PP and PS stacked data for the baseline seismic survey and the repeat survey. Compared to earlier methods based on PP data only, this method is expected to be more robust since two independent measurements are used in the computation. Due to a lack of real marine multicomponent time‐lapse seismic data sets, the methodology is tested on synthetic data sets, illustrating strengths and weaknesses of the proposed technique. Testing ten scenarios for various changes in pore pressure and fluid saturation, we find that it is more robust for most cases to use the proposed 4D PP/PS technique instead of a 4D PP amplitude variation with offset (AVO) technique. The fit between estimated and “real” changes in water saturation and pore pressure were good for most cases. On the average, we find that the deviation in estimated saturation changes is 8% and 0.3 MPa for the estimated pore pressure changes. For PP AVO, we find that the corresponding average errors are 9% and 1.0 MPa. In the present method, only 4D PP and PS amplitude changes are used in the calculations. It is straightforward to include use of 4D traveltime shifts in the algorithm and, if reliable time shifts can be measured, this will most likely further stabilize the presented method.

scholarly journals THE USE OF MACHINE LEARNING METHODS FOR BINARY CLASSIFICATION OF THE WORKING CONDITION OF BEARINGS USING THE SIGNALS OF VIBRATION ACCELERATION

2021 ◽  
pp. 15-22
Author(s):  
Ruslan Babudzhan ◽  
Konstantyn Isaienkov ◽  
Danilo Krasiy ◽  
Oleksii Vodka ◽  
Ivan Zadorozhny ◽  
...  
Keyword(s):  
Machine Learning ◽  
Binary Classification ◽  
Fractal Dimensions ◽  
Feature Space ◽  
Training Data ◽  
Supervised Machine Learning ◽  
Support Vector ◽  
Data Sets ◽  
Vibration Acceleration ◽  
K Nearest Neighbors

The paper investigates the relationship between vibration acceleration of bearings with their operational state. To determine these dependencies, a testbench was built and 112 experiments were carried out with different bearings: 100 bearings that developed an internal defect during operation and 12bearings without a defect. From the obtained records, a dataset was formed, which was used to build classifiers. Dataset is freely available. A methodfor classifying new and used bearings was proposed, which consists in searching for dependencies and regularities of the signal using descriptive functions: statistical, entropy, fractal dimensions and others. In addition to processing the signal itself, the frequency domain of the bearing operationsignal was also used to complement the feature space. The paper considered the possibility of generalizing the classification for its application on thosesignals that were not obtained in the course of laboratory experiments. An extraneous dataset was found in the public domain. This dataset was used todetermine how accurate a classifier was when it was trained and tested on significantly different signals. Training and validation were carried out usingthe bootstrapping method to eradicate the effect of randomness, given the small amount of training data available. To estimate the quality of theclassifiers, the F1-measure was used as the main metric due to the imbalance of the data sets. The following supervised machine learning methodswere chosen as classifier models: logistic regression, support vector machine, random forest, and K nearest neighbors. The results are presented in theform of plots of density distribution and diagrams.

scholarly journals Extracting horizon surfaces from 3D seismic data using deep learning

Geophysics ◽  
2020 ◽  
Vol 85 (3) ◽  
pp. N17-N26 ◽  
Author(s):  
Valentin Tschannen ◽  
Matthias Delescluse ◽  
Norman Ettrich ◽  
Janis Keuper
Keyword(s):  
Multiple Scales ◽  
Training Data ◽  
Data Sets ◽  
Seismic Image ◽  
Interpretation Process ◽  
Local Window ◽  
Subsurface Geometry ◽  
3D Seismic Data ◽  
Segmentation Problem ◽  
Practical Level

Extracting horizon surfaces from key reflections in a seismic image is an important step of the interpretation process. Interpreting a reflection surface in a geologically complex area is a difficult and time-consuming task, and it requires an understanding of the 3D subsurface geometry. Common methods to help automate the process are based on tracking waveforms in a local window around manual picks. Those approaches often fail when the wavelet character lacks lateral continuity or when reflections are truncated by faults. We have formulated horizon picking as a multiclass segmentation problem and solved it by supervised training of a 3D convolutional neural network. We design an efficient architecture to analyze the data over multiple scales while keeping memory and computational needs to a practical level. To allow for uncertainties in the exact location of the reflections, we use a probabilistic formulation to express the horizons position. By using a masked loss function, we give interpreters flexibility when picking the training data. Our method allows experts to interactively improve the results of the picking by fine training the network in the more complex areas. We also determine how our algorithm can be used to extend horizons to the prestack domain by following reflections across offsets planes, even in the presence of residual moveout. We validate our approach on two field data sets and show that it yields accurate results on nontrivial reflectivity while being trained from a workable amount of manually picked data. Initial training of the network takes approximately 1 h, and the fine training and prediction on a large seismic volume take a minute at most.

scholarly journals Image Augmentation based on GAN deep learning approach with Textual Content Descriptors

2021 ◽  
Vol 3 (3) ◽  
pp. 210-225
Author(s):  
Judy Simon
Keyword(s):  
Computer Vision ◽  
Synthetic Data ◽  
Original Data ◽  
Generative Models ◽  
Training Data ◽  
Generative Adversarial Networks ◽  
Data Sets ◽  
Biological Vision ◽  
Adversarial Networks ◽  
Textual Content

Computer vision, also known as computational visual perception, is a branch of artificial intelligence that allows computers to interpret digital pictures and videos in a manner comparable to biological vision. It entails the development of techniques for simulating biological vision. The aim of computer vision is to extract more meaningful information from visual input than that of a biological vision. Computer vision is exploding due to the avalanche of data being produced today. Powerful generative models, such as Generative Adversarial Networks (GANs), are responsible for significant advances in the field of picture creation. The focus of this research is to concentrate on textual content descriptors in the images used by GANs to generate synthetic data from the MNIST dataset to either supplement or replace the original data while training classifiers. This can provide better performance than other traditional image enlarging procedures due to the good handling of synthetic data. It shows that training classifiers on synthetic data are as effective as training them on pure data alone, and it also reveals that, for small training data sets, supplementing the dataset by first training GANs on the data may lead to a significant increase in classifier performance.

Surface-consistent matching filters for time-lapse seismic processing

Geophysics ◽  
2013 ◽  
Vol 78 (5) ◽  
pp. M29-M41 ◽  
Author(s):  
Mahdi H. Almutlaq ◽  
Gary F. Margrave
Keyword(s):  
Least Squares ◽  
Synthetic Data ◽  
Spectral Ratio ◽  
Time Lapse ◽  
Mean Value ◽  
Baseline Survey ◽  
Data Sets ◽  
Data Set ◽  
Near Surface ◽  
The Time Domain

We evaluated the concept of surface-consistent matching filters for processing time-lapse seismic data, in which matching filters are convolutional filters that minimize the sum-squared error between two signals. Because in the Fourier domain a matching filter is the spectral ratio of the two signals, we extended the well-known surface-consistent hypothesis such that the data term is a trace-by-trace spectral ratio of two data sets instead of only one (i.e., surface-consistent deconvolution). To avoid unstable division of spectra, we computed the spectral ratios in the time domain by first designing trace-sequential, least-squares matching filters, then Fourier transforming them. A subsequent least-squares solution then factored the trace-sequential matching filters into four operators: two surface-consistent (source and receiver) and two subsurface-consistent (offset and midpoint). We evaluated a time-lapse synthetic data set with nonrepeatable acquisition parameters, complex near-surface geology, and a variable subsurface reservoir layer. We computed the four-operator surface-consistent matching filters from two surveys, baseline and monitor, then applied these matching filters to the monitor survey to match it to the baseline survey over a temporal window where changes were not expected. This algorithm significantly reduced the effect of most of the nonrepeatable parameters, such as differences in source strength, receiver coupling, wavelet bandwidth and phase, and static shifts. We computed the normalized root mean square difference on raw stacked data (baseline and monitor) and obtained a mean value of 70%. This value was significantly reduced after applying the 4C surface-consistent matching filters to about 13.6% computed from final stacks.

Simultaneous multivintage time-shift estimation

Geophysics ◽  
2009 ◽  
Vol 74 (5) ◽  
pp. V109-V121 ◽  
Author(s):  
Ehsan Zabihi Naeini ◽  
Henning Hoeber ◽  
Gordon Poole ◽  
Hamid R. Siahkoohi
Keyword(s):  
Continental Shelf ◽  
Least Squares ◽  
Least Squares Method ◽  
Synthetic Data ◽  
Time Lapse ◽  
Time Shift ◽  
Data Sets ◽  
Data Set ◽  
Time Alignment ◽  
Multiple Repeat

Time-shift estimation is a key step in seismic time-lapse processing as well as in many other signal-processing applications. We consider the time-shift problem in the setting of multiple repeat surveys that must be aligned consistently. We introduce an optimized least-squares method based on the Taylor expansion for estimating two-vintage time shifts and compare it to crosscorrelation. The superiority of the proposed algorithm is demonstrated with synthetic data and residual time-lapse matching on a U. K. continental shelf data set. We then discuss the shortcomings of cascaded time alignment in multiple repeat monitor surveys and propose an approach to estimate simultaneous multivintage time shifts that uses a constrained least-squares technique combined with elements of network theory. The resulting time shifts are consistent across all vintages in a least-squares sense, improving overall alignment when compared to the classical flow of alignment in a cascaded manner. The method surpasses the cascaded approach, as noted with sample synthetic and three-vintage U. K. continental shelf time-lapse data sets.

Sign in / Sign up

Export Citation Format

Share Document