ML@GT Presents Eight Papers at SEG2018

Machine Learning Center at Georgia Tech Professor Ghassan AlRegib led the way with six of eight Georgia Tech papers published at the Society of Exploration Geophysicists (SEG) 2018., Oct. 14-19 in Anaheim, Calif.

The conference is a premier conference for AlRegib and his colleagues at Georgia Tech’s Center for Energy and Geo Processing (CeGP). The conference featured more than 1,080 presentations, 22 postconvention workshops, and more.

Georgia Tech’s papers focused on the applications of signal processing and machine and deep learning in seismic processing and interpretation.

“SEG is a great opportunity to bring together machine learning and geophysics. It is a true testament of ML@GT’s collaborative spirit and machine learning’s ability to be implemented across a broad range of topics,” said AlRegib.

For information on Georgia Tech’s presented papers please see below.

1. Title: Towards Understanding Common Features Between Natural and Seismic Images

Authors: M. A. Shafiq, M. Prabushankar, H. Di, and G. AlRegib

In this paper, the authors propose an unsupervised learning framework that aims at evaluating the applicability of the broad domain knowledge from natural images and videos in assisting seismic interpretation, such as seismic attributes, structural automation, and seismic image processing.

Read More 


2. Title: Petrophysical Property Estimation from Seismic Data Using Recurrent Neural Networks

Authors: M. Alfarraj and G. AlRegib

Reservoir characterization involves the estimation of petrophysical properties from well-log data and seismic data. Estimating such properties is a challenging task due to the non-linearity and heterogeneity of the subsurface. Researchers propose an algorithm for property estimation in seismic data using recurrent neural networks.

Read More:


3. Title: Learning to Label Seismic Structures with Deconvolution Networks and Weak Labels

Authors: Y. Alaudah, S. Gao, and G. AlRegib

Recently, there has been increasing interest in using deep learning techniques for various seismic interpretation tasks. However, unlike shallow machine learning models, deep learning models are often far more complex and can have hundreds of millions of free parameters. AlRegib and his co-authors show how automatically-generated weak labels can be effectively used to overcome this problem and train powerful deep learning models for labeling seismic structures in large seismic volumes.

Read More:


4. Title: Real-time seismic-image interpretation via deconvolutional neural network

Authors: H. Di, Z. Wang, and G. AlRegib

Seismic interpretation is now serving as a fundamental tool for depicting subsurface geology and assisting activities in various domains, such as environmental engineering and petroleum exploration. This study proposes implementing the deconvolutional neural network (DCNN) for the purpose of real-time seismic interpretation, so that all the important features in a seismic image can be identified and interpreted both accurately and simultaneously.

Read More:


5. Title: Why using CNN for seismic interpretation? An investigation

Authors: H. Di, Z. Wang, and G. AlRegib

Three-dimensional seismic interpretation plays a key role in robust hydrocarbon exploration and production of subsurface reservoirs. With the dramatic growing size of 3D seismic surveys, however, manually interpreting a seismic volume turns to be even more challenging. In this study, researchers first apply two most popular neural network frameworks, the multi-layer perceptron (MLP) network and the convolutional neural network (CNN), to the problem of seismic salt-body delineation and compare their performance. They then investigate two factors that contribute to the better performance of the CNN framework in understanding seismic signals and identifying the important seismic structures.

Read More:


6. Title: Patch-level MLP classification for improved fault detection

Authors: H. Di, M. A. Shafiq, and G. AlRegib

Fault detection and interpretation have been one of the routine tools used for subsurface structure mapping and reservoir characterization from three-dimensional (3D) seismic data. With the recent developments in machine learning and big data analysis, this study proposes an innovative method for efficient seismic fault detection based on semi-supervised classification of multiple attribute patches through the popular multi-layer perceptron (MLP) technique.

Read More:


7. Title: Automatic microseismic event detection using constant false alarm rate processing in time-frequency domain

Authors: A. Raj, J.H. McClellan, N. Iqbal, A.A Al-Shuhail, and S.I. Kaka

Detecting and monitoring microseismic events using surface sensors in unknown noise scenarios and low signal-to-noise ratio conditions is a challenging problem. Researchers propose a scheme for reliable automatic detection of microseismic events based on 2D Constant False Alarm Rate (CFAR) processing in the time-frequency (TF) domain, along with an efficient 2D filtering implementation of the 2D CFAR algorithm.

Read More:


8. Title: Sum-Rate Maximization for Wireless Seismic Data Acquisition Systems

Authors: A.Othman, W. Mesbah, N. Iqbal, S. Al-Dharrab, A. Muqaibel, and G. Stuber

In this paper, the authors consider the problem of maximizing the information theoretic sum-rate in a wireless geo-seismic acquisition system.

Read More:

Related Media

Click on image(s) to view larger version(s)

  • SEG 2018 was held Oct. 14-19 in Anaheim, Calif.

For More Information Contact

Allie McFadden

Communications Officer