CoreLithoClassifier

A demonstration project integrating geology, machine learning, and image processing to classify core samples based on synthetic geochemical data and core imagery.

Project Objective

This project simulates how core imagery and associated geochemical data can be used together to build lithology classifiers. It mimics a simplified version of what core scanning AI systems aim to achieve.

Features

• Synthetic core imagery generator for Mafic, Ultramafic, and Felsic lithologies
• Geochemical data simulation (Ni, Cu, Fe, S)
• Image preprocessing using OpenCV and PIL
• Feature extraction (color histograms, edge detection, GLCM texture)
• Geochemistry-based PCA and clustering
• CNN model to classify lithology directly from images

Tech Stack

• Python
• NumPy, Pandas, Matplotlib
• Scikit-learn
• OpenCV, Pillow
• TensorFlow/Keras (for CNNs)

File Structure

CoreLithoClassifier/
├── data/
│   ├── sample_core_images/       # Synthetic core images
│   └── geochemistry.csv          # Synthetic geochemical dataset
├── notebooks/                    # Jupyter notebooks
├── output/
│   └── figures/                  # Output plots
├── src/                          # Python scripts for modeling & utilities
├── README.md
└── requirements.txt

Next Steps

• Explore geochemical data (PCA, clustering)
• Train CNN on core images
• Combine features for hybrid lithology prediction

---

Disclaimer: All data used in this project is synthetic and intended for demonstration purposes only.