High-resolution classification of subsurface features from directional image gathers, based on Principal Component Analysis and Deep Learning methods

Abstract

Specular/diffraction imaging has proven to be an attractive approach to providing high-resolution subsurface images containing different types and scales of continuous and discontinuous geometrical objects. However, there is a strong need in exploration and production to continue investigating a direct method for detecting and characterizing fractures and other small-scale reservoir heterogeneities that would impact understanding of the reservoir’s producibility. Seismic response from such small-scale structural and lithological elements in the subsurface is encoded in the diffraction part of the total wavefield. As part of the ongoing effort to enhance procedures for interpreting/classifying directivity-driven image data, we present a Deep Learning (DL) approach to this challenging task. Preliminary results indicate great promise in using this emerging technology. It enables structural features, such as small-scale faults pinchouts and fractures, to be more readily and reliably localized, identified and characterized on the diffraction images.

AAPG Datapages/Search and Discovery Article #90352 © 2019 AAPG Latin America and Caribbean Region Geoscience Technology Workshop, Rio de Janeiro, Brazil, June 13-14, 2019