Project Info


Automated Lithology Prediction from Digital Outcrop Models Using Machine Learning

Zane Jobe | zanejobe@mines.edu and Wendy Fisher | wfisher@mines.edu

Drone-derived photogrammetry has become a ubiquitous data-collection technique for field geologists (in addition to many other fields, including construction and mining). However, currently there are only limited tools for interrogating drone-derived 3D digital outcrop models to provide information about lithology proportions and stacking patterns. This project focuses on supervised-learning approaches to lithology classification from digital outcrop models that will allow rapid generation of quantitative data on spatial heterogeneity of rock type and layering. Preliminary work on this topic has shown that RGB image classification and the texture of the 3D mesh can be used to classify lithology, and this project will further evaluate those techniques as well as incorporate hyperspectral and other sensor data.

Elucidate the interdisciplinary nature of the project

This project utilizes computer-science methods to solve geologic problems. Faculty in both the geology (Zane Jobe, Gabe Walton) and computer-science departments (Wendy Fisher) as well as a research associate with data science expertise (Ross Meyer) will work together with the MURF recipient(s) to bring cutting-edge computer-science solutions to current geologic problems that have societal relevance and industry applicability.

More Information

For previous work on this project, see the results from a 2 day hackathon:
Presentation: https://docs.google.com/presentation/d/1iFGPUWKGjMO1G03XbT0NZ6DECdnsnOmK8iWwxezjmcE/edit?usp=sharing
Visualization: https://vimeo.com/270994566?utm_source=email&utm_medium=vimeo-cliptranscode-201504&utm_campaign=28749
GitHub repo: https://github.com/rgmyr/OutcropsGeeWhiz
Example of 3D outcrop model: https://sketchfab.com/models/fe85a8ae79bb465eb6370d046616767f

Grand Engineering Challenge: Enhance virtual reality

Student Preparation


Qualifications

Student #1- Basic understanding of geology is a must, but all earth-science fields encouraged (Geophysics, Pet. Eng., etc.). Preferred (but not mandatory) prerequisistes include GEGN 101 (Intro to Geology), GEGN 204 (Geologic Processes/Principles), and GEOL 314 (Sedimentology/Stratigraphy).

Student #2-Basic understanding of python is a must. Experience in software development, machine learning, and geology preferred but not mandatory. Preferred (but not mandatory) prerequisistes include CSCI 101 (Intro to Comp. Sci.) and CSCI 303 (Intro to Data Science), and CSCI 470 (Machine Learning)

Time Commitment

20 hours/month

Skills/Techniques Gained

Both students will learn how to design, implement, and complete a research project, including releasing open-source software and writing a peer-reviewed publication. Depending on the student’s focus, specific technical skills will include knowledge of stratigraphic architecture and sedimentary geology, drone flying, digital outcrop creation using drone-based photos and Structure-from-Motion software, python-based software design, GitHub version control, image classification, supervised machine learning methods, and data visualization.

Mentoring Plan

Zane and Gabe will provide mentoring in digital outcrop model construction and geologic interpretation. Ross and Wendy will provide mentoring on supervised machine learning approaches, image classification, and data visualization. We will have monthly meetings to discuss progress with the MURF receipient(s).