| --- |
| license: mit |
| --- |
| # 🪨 Dataset Description: `segments_with_geochem.csv` |
|
|
| ## 🧾 Overview |
| This dataset is a **merged collection** linking **image segments of drill core trays** with their corresponding **lithological and geochemical information**. |
| Each row represents one image segment (a small interval of core depth) combined with a specific geochemical measurement. |
|
|
| It contains **17,992 rows** and **31 columns**, integrating three main data sources: |
| 1. **Segmented core images** generated from hyperspectral core tray processing. |
| 2. **Lithology intervals** from SARIG drillhole logging data. |
| 3. **Geochemical assay data** matched by drillhole and depth range. |
|
|
| --- |
|
|
| ## 🧱 Structure and Key Columns |
|
|
| | Category | Example Columns | Description | |
| |-----------|------------------|--------------| |
| | **Image Segment Metadata** | `filename`, `drillhole_number`, `core`, `segment`, `depth_m` | Derived from processed images (e.g. `9955_5_9_9.15.jpg`). Each represents a core segment at a specific depth within a drillhole. | |
| | **Lithology Data (SARIG)** | `LOG_NUMBER`, `LOGGING_ORGANISATION`, `QUALIFICATION`, `LOGGING_DATE`, `major_lithology`, `DESCRIPTION` | Imported from SARIG drillhole lithology logs. These describe rock type, logging geologist, and confidence of lithological interpretation. | |
| | **Spatial Coordinates** | `EASTING_GDA2020`, `NORTHING_GDA2020`, `LONGITUDE_GDA2020`, `LATITUDE_GDA2020`, `ZONE_GDA2020` | Georeferencing information for each drillhole site. | |
| | **Geochemistry Data** | `chem_code`, `value`, `unit`, `chem_method_code`, `chem_method_desc` | Chemical element concentrations and assay method details. For example, `chem_code = 'Cu'`, `value = 85`, `unit = 'ppm'`, `chem_method_code = 'IC4M'`. | |
| | **Depth Range (Lithology/Assay)** | `depth_from_m`, `depth_to_m` | Defines the downhole interval associated with the lithology or geochemical measurement. | |
|
|
| --- |
|
|
| ## 🧬 How It Was Made |
|
|
| ### 1. Core Segmentation |
| Core tray images were processed into **individual core segments** using depth data. |
| Each segment was saved as an image file named using this pattern: |
| [drillhole_number]\_[core_number]\_[segment_number]\_[depth_m].jpg |
| |
| Example: `9955_5_9_9.15.jpg` → Drillhole 9955, Core 5, Segment 9, Depth 9.15 m. |
|
|
| --- |
|
|
| ### 2. Lithology Mapping |
| Each image segment was matched to **SARIG lithology logs** (`sarig_dh_litho_exp.csv`), using: |
| drillhole_number matches AND |
| depth_from_m <= segment_depth <= depth_to_m |
|
|
| This allowed each segment to inherit lithological attributes such as `major_lithology` and `DESCRIPTION`. |
|
|
| --- |
|
|
| ### 3. Geochemical Integration |
| The merged segment–lithology dataset was further joined with **SARIG assay data**. |
| Matching was again done by `drillhole_number` and depth overlap. |
| Each segment now includes one or more geochemical assay results (e.g., Cu, Fe, Au) with their `chem_method_code` and measurement units (commonly `ppm`). |
|
|
| --- |
|
|
| ## 📊 Purpose and Use |
| This dataset is designed for **data integration and machine learning** applications that connect visual and chemical rock properties. |
| It can be used for: |
| - Linking **visual mineralogical features** in core images to **geochemical compositions**. |
| - Building **ML models** to predict chemistry or lithology from imagery. |
| - Performing **cross-domain correlation** between **NVCL hyperspectral data** and **laboratory assays**. |
|
|
| Each record connects what’s *seen* in the rock (image/lithology) to what’s *measured* in the lab (chemistry). |
|
|
| --- |
|
|
| ## 🧪 Notes |
| - The **`value`** column represents assay readings, often in **ppm (parts per million)**. |
| - The **`unit`** column confirms the measurement type (e.g., `ppm`, `%`, etc.). |
| - The dataset integrates spatial, spectral, and chemical data — making it suitable for **multi-modal mineral systems analysis**. |
|
|
