--> Deep Convolutional Neural Networks for Seismic Salt-Body Delineation

AAPG ACE 2018

Datapages, Inc.Print this page

Deep Convolutional Neural Networks for Seismic Salt-Body Delineation

Abstract

Salt-bodies are important subsurface structures with significant implications for hydrocarbon accumulation and sealing in offshore petroleum reservoirs, and accurate salt-body imaging and delineation is now greatly facilitated with the availbility of 3D seismic surveying. However, considering the growing complexity of seismic data in both size and resolution, the efficiency of interpreting a salt-body increasingly relies on the development of powerful computational interpretation tools that are capable of mimicking an experienced interpreter’s intelligence. In recent years, the machine learning has been successful for image/video understanding in various disciplines and is attracting more and more attentions from the petroleum industry. Geoscientists desire to explore the massive seismic data in more intelligent ways to extract more information for better understanding the subsurface reservoirs. This study implements the emerging convolutional neural network (CNN) for the purpose of salt-body delineation from 3D seismic data, which is superior in two ways compared to the traditional sample-based multi-attribute classification schemes. First, the CNN takes into account the local seismic patterns for defining and learning the target salt-body features, so that the coherent noises and processing artifacts of distinct patterns can be effectively identified and excluded. Second, the CNN builds an opimal mapping relationship between the seismic signals and the salt-bodies directly from the reflection amplitude, which avoids the process of manual attribute selection where interpreter bias might be introduced. The added values of such CNN-based classification is demonstrated through applications to the synthetic SEG-SEAM dataset, which is featured with a complex intrusive in a folded Tertiary basin and challenges the existing salt-body interpretation tools. The preliminary results not only show good match between the detection and the original seismic images particularly in the zones where the reflection is weak and not discernable to the existing seismic attributes, but also indicate the great potential for applying the CNN tool to computer-aided extraction of other important seismic objects (e.g., faults).