Computer Vision on Eigenform AI Geo Tooltips

AI Image Analysis for Ore Texture, Grain Size and Mineral Mapping

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Textural analysis, grain size analysis, and mineral mapping run on the same automated-mineralogy and hyperspectral imaging stack: TESCAN TIMA, Zeiss Mineralogic, and Corescan for core-scale hyperspectral work.
These platforms increasingly use CNN-based classifiers rather than fixed spectral lookup tables to resolve grain boundaries and mixed-pixel mineralogy.
Output feeds two downstream uses directly: comminution/liberation modeling (how finely you need to grind to liberate value minerals) and geometallurgical domaining.
This is a mature, commercially deployed capability (ai_relevant: Yes across all three) — not a research prototype.
Malvern Panalytical’s Morphologi line is a relevant alternative specifically for particle-scale grain-size distribution work outside the SEM-based platforms.

TL;DR

Run polished samples or drill core through an automated SEM mineralogy platform (TIMA, Mineralogic) or hyperspectral core scanner (Corescan) to get quantitative texture, grain-size, and mineral-map data with built-in ML-based classification, rather than manual petrographic description.

AI-Assisted Automated Mineralogy: TIMA, MLA & QEMSCAN

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

TIMA, MLA, and QEMSCAN are all SEM-based automated mineralogy platforms — they scan a polished sample, fire an electron beam, and classify every mineral grain from its X-ray spectrum, at rates up to ~100,000 grains/hour.
All three now lean on machine-learning classifiers under the hood for particle/phase identification, not just fixed spectral lookup tables.
MLA is the original platform (Thermo Fisher/FEI); QEMSCAN and Zeiss Mineralogic are close alternatives, with Mineralogic explicitly marketing AI-based deep-learning classification.
TESCAN TIMA is the newer entrant and increasingly the default choice for high-throughput exploration and geometallurgy programs.
Output feeds directly into liberation analysis, grain-size distribution, and the clay/hardness/recovery ML models used elsewhere in your geometallurgy workflow.

TL;DR

Send polished sample mounts through an automated SEM mineralogy platform (TIMA, MLA, Mineralogic, or QEMSCAN) rather than manual point-counting — the instrument’s built-in ML classifier turns raw X-ray spectra into quantitative mineral, liberation, and grain-size data in hours instead of weeks.

Automating Ore Sampling with Sensors and AI

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Shovel-face, automated, and blasthole/RC sampling are all being upgraded with sensor packages that reduce manual handling and speed up grade turnaround.
Motion Metrics ShovelMetrics and MineSense’s shovel sensor line use AI/computer vision to estimate fragmentation and grade at the dig face in real time.
Scott Automation’s Rocklabs line (AMS Prep, RoboPrep Elite) automates crushing, splitting, and pulverizing so cross-belt or lab autosamplers can feed inline XRF without manual prep.
IMDEX BLASTDOG specifically instruments blasthole/RC sampling, capturing sensor-while-drilling data as the hole is drilled.
Blasthole/RC sampling AI is the least mature of the three (Emerging) — most of the intelligence currently sits in the downstream XRF/assay step rather than the sampling mechanism itself.

TL;DR

Replace manual scoop-and-bag sampling with sensor-equipped shovels (grade/fragmentation AI), robotic sample-prep lines feeding inline XRF, and instrumented blasthole rigs like BLASTDOG — cutting the lag between digging ore and knowing what’s in it.

Bench Slope Design

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Bench slope design still runs on classical finite-element and limit-equilibrium software — RocScience’s Slide2/RS2/RS3 and GeoStudio remain the industry standard.
The AI layer is emerging and upstream of the design software: machine-learning models predicting rock-mass properties (dip, dip-direction, strength) feed faster, more automated slope-design inputs.
Don’t expect an “AI slope designer” product — expect AI-derived inputs plugged into the same trusted geotechnical engines you already use.

TL;DR

AI-assisted bench slope design means using machine-learning models to generate structural and rock-mass inputs (like dip/dip-direction from drillhole or image logs) faster, then running the actual factor-of-safety analysis in established tools like RocScience Slide2/RS2/RS3 or GeoStudio, same as always.

Blasthole Logging

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Hexagon Drill Assist puts AI directly into the drilling algorithm, automatically managing drill parameters from rig-sensor data as it drills — no manual parameter entry.
Operators report training time dropping from years to about 15 minutes with Drill Assist, alongside real gains in fragmentation, ore dilution reduction, and energy use per metre.
Fully automated computer-vision lithology/cuttings logging is still research-stage — sensor-while-drilling systems from IMDEX and Epiroc are the more mature adjacent capability.

TL;DR

AI blasthole logging today mostly means AI-driven drilling itself — Hexagon Drill Assist manages drill parameters automatically from live rig-sensor data, capturing consistent, high-quality blasthole logs as a byproduct, while automated cuttings/lithology vision-based logging remains a research direction rather than a deployed product.

Blasting Evaluation

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Post-blast evaluation splits into two AI-driven tasks: tracking where the ore actually moved (Hexagon Blast Movement Intelligence) and measuring fragmentation size (Split-Desktop, WipFrag).
Hexagon BMI generates a “Muckpile Block Model” — a post-blast update to your grade-control block model — without requiring personnel to walk the muckpile to place or retrieve monitors.
Fragmentation-analysis tools use image-processing/computer-vision techniques on muckpile photos to estimate particle size distribution instead of manual sieving.
This is a genuinely mature AI application — both categories have years of field deployment behind them, not just pilot studies.

TL;DR

Evaluate blast performance with two complementary AI tools: Hexagon Blast Movement Intelligence for tracking ore movement into a muckpile block model, and image-analysis fragmentation tools like Split-Desktop or WipFrag for particle size distribution.

Core Photography

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

High-resolution automated core photography is the raw data source that most downstream AI core-analysis tools depend on.
Corescan (now Epiroc), Datarock Core, and Minalyze all offer automated core photography systems designed to feed AI extraction pipelines.
This is a “Yes” — production-grade, already deployed across major mining companies.
Get the photography pipeline right first: every AI fracture/lithology/mineralogy tool downstream is only as good as the image quality and consistency it’s trained/run on.

TL;DR

Set up automated, standardized high-resolution core photography (via Corescan, Datarock, or Minalyze hardware) as the foundation — it’s the input every downstream AI logging, fracture-detection, and mineralogy tool depends on.

Flotation

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Flotation is one of the most mature applications of industrial AI in mineral processing — this isn’t speculative.
Imubit’s closed-loop AI uses reinforcement learning to write optimal setpoints directly to the plant’s existing control system, not just recommend them to an operator.
Reported gains: up to ~15% recovery improvement and ~20% reagent-use reduction — real deployments, not lab results.
Froth-camera vision models (including ConvLSTM-based research systems) are the sensing layer that feeds these controllers real-time froth texture and bubble-size data.

TL;DR

Feed a froth camera’s real-time image stream and plant sensor data (reagent dosage, pulp density, pH, airflow) into a closed-loop reinforcement-learning controller like Imubit, and let it write setpoints back to your DCS in real time instead of relying on operators to react to trends.

Fracture Spacing AI

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Fracture spacing/frequency logging is one of the best-validated AI applications on the whole poster — it has a published head-to-head comparison against expert human logging.
Datarock Core (part of IMDEX) uses a neural network to detect and classify fractures directly from core photography.
The Carrapateena copper-gold mine study specifically validated this against expert geotechnical logging — a real deposit, real comparison, not a synthetic benchmark.
This is a “Yes” — the clearest example on the whole list of AI matching (not just approximating) expert human judgment in a specific geotechnical task.

TL;DR

Run core photography through Datarock Core’s neural-network fracture detection model to get automated fracture spacing/frequency logs that have been validated against expert human geotechnical logging in a real copper-gold mine study.

Frag TRACK

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Orica FRAGTrack uses AI plus stereoscopic 2D/3D imaging for automated post-blast fragmentation analysis — no manual sieving or scaled photos required.
It ships in multiple deployment forms: FRAGTrack Conveyor, FRAGTrack Crusher, FRAGTrack GeoSpatial, and a newer excavator-mounted variant — covering everything from the dig face to the crusher.
WipWare WipFrag and Split-Desktop are the long-standing image-analysis alternatives if you want a lower-cost or desktop-based option.
Binocular/stereoscopic camera capture is what lets these systems handle variable lighting and material color/texture without needing a fixed reference scale in every shot.

TL;DR

Fragmentation analysis is done today with AI-driven stereoscopic image analysis — Orica’s FRAGTrack is the current market-leading deployed system, with WipFrag and Split-Desktop as established, lower-friction alternatives for teams not ready for a full sensor-network rollout.

Informed Logging

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

“Informed” logging means the geologist (or AI system) logs new core with the benefit of prior geological/geophysical model context, not in isolation.
Datarock Core and Seequent Central’s logging module both support this by surfacing existing model context alongside new imagery.
This is a “Yes” — the underlying AI core-logging technology is production-proven; what’s new is using it context-aware rather than hole-by-hole in isolation.
The practical benefit: consistency across a drill program, since every hole gets logged against the same evolving model rather than each geologist’s independent read.

TL;DR

Pair an AI core-logging tool like Datarock Core with a centralized model platform like Seequent Central so that each new hole is logged with the benefit of everything already known about the deposit, rather than being interpreted in isolation.

Ore Sorting

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Sensor-based ore sorting is a mature, high-adoption category — TOMRA and Steinert together hold over 40% of the market.
The “AI” is in the real-time classification logic that decides, particle-by-particle, whether a rock is ore or waste based on sensor signal — this has moved well beyond simple threshold rules.
TOMRA’s AI-powered advancements (OBTAIN, CONTAIN) specifically target throughput (doubling sorting capacity) and fine inclusion-detection for base metal sulphides including copper.
This is a bulk pre-concentration step — sorting waste out before it reaches the mill — which cuts downstream energy, water, and tailings volume, not just a grade-control tool.

TL;DR

Adopting AI-driven ore sorting means installing a sensor-based sorter — TOMRA or Steinert KSS are the two dominant vendors — ahead of your mill, where AI classification software decides in real time which rocks to keep and which to reject.

Ore Sorting Technologies

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

The competitive frontier in ore sorting isn’t a single sensor — it’s sensor fusion: combining XRT, optical/machine vision, laser, and induction/NIR signals on one platform.
Steinert’s KSS is explicitly built as a combined sensor system, letting you pair XRT, XRF, or NIR with color, 3D laser, and induction sensing depending on ore type.
Which sensor combination works depends heavily on what physically distinguishes your ore from waste — density and atomic number differences favor XRT; surface color/texture favors optical; conductivity differences favor induction.
AI classification software is what turns multiple simultaneous sensor streams into a single accept/reject decision per particle — this is where the real technology differentiation between vendors now sits.

TL;DR

Choosing an ore-sorting technology means matching sensor type(s) to what physically distinguishes your ore from waste, then relying on the vendor’s AI classification software to fuse those signals — Steinert’s combined KSS platform is built explicitly for multi-sensor fusion, while TOMRA offers XRT, NIR, and laser as configurable options.

Petrography

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Petrographic thin-section analysis (identifying minerals and textures under a microscope) is still overwhelmingly analyst-driven — there’s no single dominant commercial AI product yet.
Published CNN research is real and promising: concatenated convolutional neural network approaches have hit ~90% accuracy classifying rock type from plane- and cross-polarized thin-section images.
Treat this as an emerging, DIY-research capability rather than something you can buy and deploy this quarter.

TL;DR

AI-assisted petrography means training a convolutional neural network on your digitized thin-section images (plane- and cross-polarized light) to automate rock-type and mineral classification — technically proven in the literature, but not yet a packaged commercial tool.

Predicting Geotechnical & Recovery Variables in Your Block Model with ML

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

RQD, fracture frequency, point-load strength, clay content, comminution hardness (BWI/DWI), and metal recovery (Cu, Mo, Au, Ag) can all be estimated with machine learning instead of relying purely on sparse physical test coverage.
Computer-vision models (e.g. CNN-based “K-Net” pipelines) now read RQD and fracture frequency directly from core photographs, cutting manual logging variance.
Published copper-recovery models (XGBoost/Random Forest/ANN) have delivered real production gains — a Cerro Verde case study reported +6.5% Cu production from ML-driven recovery prediction.
Most of these submodels aren’t a single commercial product — they’re scikit-learn/XGBoost regressions trained on your own assay + geological + lab-test database, then interpolated into the block model like any other estimated variable.
Maturity varies: RQD/recovery models are proven in production; PLT-model prediction is still mostly a custom, in-house build (Emerging).

TL;DR

Train gradient-boosted or neural-network regressors on your existing lab-test results (RQD, PLT, BWI/DWI, clays, recovery) against geology/alteration/mineralogy features, then krige or interpolate the model’s predictions the same way you would a raw assay — giving every block an estimated geotechnical/metallurgical value instead of only the handful with physical tests.

Quick Geological Logging

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

First-pass (“quick”) geological logging — the initial lithology/alteration read on fresh core — is one of the more mature AI use cases in exploration.
Datarock Core and Seequent Central’s logging module both support AI-accelerated first-pass logging directly from core photography.
Datarock has processed over 30 million metres of core since 2020, turning what would be months of manual logging into hours.
This is a “Yes” — already deployed, production-grade AI, not a research prototype.

TL;DR

Feed high-resolution core photography into a tool like Datarock Core to get an automated first-pass lithology/structure log in hours instead of the days or weeks manual logging takes, then have your geologists review and refine rather than log from scratch.

Structural Mapping

Mon, 21 Sep 2026 00:00:00 +0800

Key Takeaways

Photogrammetric structural mapping — extracting discontinuity orientation, spacing, and trace length from pit-wall imagery — is the established standard, via Maptek’s I-Site Studio/PointStudio and the CSIRO-developed Sirovision system.
Automated discontinuity-detection algorithms are increasingly built into these platforms, reducing manual digitizing of joint sets from point clouds.
This is a commercial, deployed capability, with the AI/automation layer still emerging on top of an already-mature photogrammetry foundation.

TL;DR

AI-assisted structural mapping means capturing pit-wall point clouds via laser scan or drone photogrammetry in Maptek PointStudio (or CSIRO’s Sirovision), then using increasingly automated discontinuity-detection algorithms to extract joint orientation, spacing, and trace length instead of hand-digitizing every structure.