Loading...
Towards lithology recognition using digital outcrop models and machine learning
Chacon Buitrago, Nataly
Chacon Buitrago, Nataly
Citations
Altmetric:
Advisor
Jobe, Zane R.
Walton, Gabriel
Walton, Gabriel
Editor
Date
Date Issued
2022
Date Submitted
Collections
Research Projects
Organizational Units
Journal Issue
Embargo Expires
Abstract
In sedimentary geology, outcrop studies are used to characterize and understand depositional environments and generate analogs for subsurface characterization. Lithological classification is one of the building blocks for outcrop studies and thus should be a reproducible and time-efficient task. In recent years, geologists have collected numerous 3D outcrop models that can be used to develop reproducible lithological classification models in a time-efficient way. The goal of this study is to provide a rapid, reproducible workflow for outcrop point cloud analysis that will generate quantitative lithology data, thereby capturing the lithological variations in large areas, which eventually will decrease the time necessary to build accurate models for subsurface characterization of earth resources (e.g., hydrocarbons, CO2, water, etc.).
High-resolution point clouds of siliciclastic turbidite outcrops of the Cloridorme formation in the Gaspe Peninsula in Canada and the Point Loma Formation in San Diego, California, were used as the primary datasets for this study. These are ideal sites for methodology development, as there are clear distinctions between lithologies. Features (e.g., color, texture, slope, geometry) were extracted from the point clouds using multiple search radii. These features were inputs to machine learning algorithms (e.g., Random Forest, K-Nearest Neighbors, and Support Vector Machine) used to classify the lithology of the point cloud. Comparison of results from the classification task to manually labeled point clouds suggests that accuracies of 78%+ and 80%+ can be achieved when using this algorithm for lithology classification. These are promising results, but further work has to be done.
Associated Publications
Rights
Copyright of the original work is retained by the author.