Skip to main content Skip to secondary navigation

Machine learning for geological image analysis

Main content start

Analyzing images plays a major part in the work of geoscientists on a range of spatial scales, from interpreting regional seismic data to identifying and classifying lithofacies in thin sections. This visual information is crucial when solving many geoscience problems, including reconstructing depositional environments, paleocommunities, and tectonic history as well as understanding the quality and heterogeneity of reservoir strata. Conventional image analysis is often time consuming and costly, and frequently requires in-depth knowledge of specific geological sub-fields (e.g. carbonate geology). In principle, computer algorithms have the potential to increase the speed of image interpretation and simultaneously provide a high accuracy analysis. Recently, Deep Convolutional Neural Networks (DCNN) have been increasingly used in geosciences to automate the identification and interpretation of petrographic images, core photos, and seismic data. To date, however, only a few studies have applied the DCNN method to the classification and identification of carbonate rock constituents in static or live petrographic images. Because carbonate rocks serve an important role as aquifers and hydrocarbon reservoirs, it is imperative that automated image interpretation is explored as a potential tool to improve the efficiency and accuracy of carbonate reservoir characterization. This robust image classification approach will be used to classify geologic images on the well bore and basin scales as well.

Graph Diagram

Related People