Key Takeaways
- Hyperspectral scanning comes in two flavors: handheld point spectrometers (fast spot checks) and full-core imaging scanners (slower, exhaustive and image-based).
- The core technology, VIS-NIR-SWIR reflectance spectroscopy, is mature and commercial. The “AI” part is mostly the automated mineral-classification layer on top of the raw spectra.
- Malvern Panalytical’s ASD TerraSpec line handles point measurement. Corescan’s Hyperspectral Core Imager handles full-core scanning.
- Treat this as a data-acquisition layer, not a standalone AI product. The value shows up downstream in alteration and clay models.
TL;DR
Choose point or imaging hyperspectral scanning based on the question: a handheld spectrometer for quick spot checks, a core imager for continuous alteration maps. Then feed the spectra into a classification or regression model to get mineral abundances. Budget for the interpretation model separately from the hardware.
What Is Hyperspectral Scanning?
Every mineral reflects and absorbs light slightly differently across the visible and infrared spectrum. Clays, micas, carbonates and iron oxides each have characteristic absorption features. A hyperspectral sensor records hundreds of narrow wavelength bands, so it captures those features in detail.
Hyperspectral scanning uses that signal to identify minerals without cutting or dissolving the rock. In exploration and geometallurgy, the term covers two distinct workflows that often get lumped together on planning diagrams. One measures single points. The other images entire trays of core. They answer different questions, so it helps to separate them.
How Do I Choose Point or Scan Hyperspectral With AI?
The decision comes down to coverage, then to how you turn spectra into minerals.
Point spectrometers for spot checks
A geologist in the field or core shed points a handheld spectrometer at a rock face or core sample. Common instruments are the ASD TerraSpec 4 and its now-discontinued predecessor, the TerraSpec Halo. The instrument matches the spectrum against a library of thousands of reference minerals in seconds. This is fast and cheap per sample, but sparse: you only see what you point at. Our post on the portable spectrometer covers this workflow in more depth.
Imaging scanners for full-core hyperspectral scanning
The second workflow runs an entire drill core past a sensor. Corescan’s Hyperspectral Core Imager (HCI-3 and HCI-4, now sold through Epiroc) combines a VNIR-SWIR sensor with a visual camera and laser profiler. It collects roughly 800,000 spectral samples per metre. That density is what makes downstream AI useful. A sparse handheld dataset is rarely enough training data for a classifier, but a dense core scan is. For the wider range of core scanning systems, including XRF and thermal infrared, see our separate guide.
Classifying hyperspectral scanning data with AI
Neither system is “AI” out of the box; they are spectrometers. The AI angle, and where this is still emerging rather than mature, is the interpretation layer. It turns raw reflectance curves into mineral percentages, alteration intensity or clay content. For example, CSIRO’s MyLogger applies neural-net interpretation to HyLogger data. Other regression pipelines turn Corescan output into inputs for high-resolution clay models.
Budgeting for a Hyperspectral Scanning Program
If you are evaluating this stack, budget separately for the scanner hardware and the interpretation model. They are sold and built differently. Hardware is a capital or service cost, while the model needs geologist time, lab calibration samples and ongoing validation.
Many programs use both approaches. A handheld unit screens field samples and checks scanner results, while an imaging system covers every metre of priority holes. However, both measure only the rock surface. Therefore, calibrate spectral outputs against XRD or automated mineralogy on a subset of samples before trusting them in a block model.
FAQs
What is the difference between point and imaging hyperspectral scanning?
Point spectrometers measure one small spot at a time and give an instant mineral match. Imaging scanners capture continuous spectra across whole core trays. Point tools suit spot checks, while imaging suits systematic mapping.
Which minerals can hyperspectral scanning identify?
VNIR-SWIR sensors are strongest for clays, micas, chlorite, carbonates, sulphates and iron oxides. Many anhydrous silicates, such as quartz and feldspar, need thermal infrared instead. Sulphides are hard to identify with reflectance spectroscopy.
Does hyperspectral scanning need AI?
Not for basic mineral identification, which uses spectral library matching. AI becomes valuable when you want quantitative abundances, alteration intensity or clay content across millions of spectra. It also links spectral data to assays and metallurgical tests.
Is hyperspectral scanning worth it for a small drill program?
Often a handheld spectrometer is enough for a small program. Imaging scanners pay off when you have many holes and need continuous, consistent logs. Service providers let you scan core without buying the hardware.
Try It With Geocluster
Turning raw spectral scans into a defensible geomet model means wiring together instrument output, reference spectral libraries and a classification model. That multi-source research workflow is what Geocluster is built to orchestrate. Read how the Geocluster Research Harness works if you are stitching hyperspectral data into your exploration pipeline.