English Russian


Automatic Recovery and Classification of Seismic Wavefields from Prestack Seismic with Deep Learning

October 23, 2018
This presentation is now available for viewing.

Presented by:  Sandra Allwork, Business Development Manager, Emerson E&P
Featured Domain:  Processing and Imaging
Featured Product:  EarthStudy 360


For comments or questions, please contact Sandra Allwork.


This presentation describes the most recent evolution of Emerson’s revolutionary utilization of the Local Angle Domain to characterize subsurface features from seismic data.  Knowing that the EarthStudy 360 system enables us to decompose the recorded seismic wavefield into full-azimuth directivity components, we will demonstrate the use of Principle Component Analysis (PCA) with its inherent compression, to derive principle directivities.  
Ultimately, we will use the power of Deep Learning  to classify these directivities  into geometrical, indeed geological  features such as faults, plus other identifiable components such as ambient noise or acquisition footprint. This method presents a more reliable method for separating these components, thus producing clearer stack images of the subsurface. They show superior results over previously unrivaled diffraction weighted stacks.


_IWM4715_sm.jpgSandra Allwork has a BSc in Geology & Geophysics from the University of Durham in the UK.  Before joining Paradigm, she worked for seven years in the seismic processing service sector. Sandra  is now Director of Technical Services at Emerson E&P Software in Europe, and has  20+ years’ experience with the Paradigm portfolio. Over the years, Sandra has worked with customers in Europe and the Middle East to optimize their velocity model building and imaging workflows, incorporating the latest advances in anisotropy and full-azimuth analysis.