Mines are full of cameras. Longwall faces, conveyors, intersections — operators rely on them constantly.
But mines are also full of dust, water spray, and stone dust. Cameras get dirty. And when they do, nobody notices until something goes wrong.
Most sites already have spray systems to clean their cameras. But they run on timers — fixed intervals regardless of whether the lens is clean or caked in dust. That means cameras get over-cleaned when they don't need it (wasting water and wear), and under-cleaned when they do (leaving operators with a degraded image right when it matters most).
We built a system that replaces the timer with intelligence.
JMB Automation developed a dirty camera detection system for Kestrel Coal's underground longwall operation. It uses a custom trained machine learning classification model running alongside the existing camera network to continuously assess image quality in real time. When a camera's view degrades beyond threshold the system doesn't just flag it. It triggers the cleaning spray automatically, only when it's actually needed.
This is what we mean by wrap-around automation at JMB Automation. It's not one piece of the puzzle — it's the full picture:
🔹 Purpose-built AI — the model was trained and validated specifically on underground coal mine camera imagery, not generic datasets. The environment is unique — dust isn't noise, it's a constant — so the model has to be built for it.
🔹 Control system integration — the AI feeds directly into the existing PLC and Citect SCADA architecture. We're not just monitoring and providing data — we're making decisions and taking automated action. No separate platform, no manual intervention.
🔹 Shearer-aware intelligence — the system tracks the shearer's MG-TG run, so it knows which cameras matter most at any given moment and prioritises accordingly.
🔹 Unified operations dashboard — all of our smart products have the ability to link back to a single platform giving operations a complete view of their automations and data in one place. Dirty camera detection sits alongside our other intelligent systems as part of a connected picture, not an isolated tool.
This isn't a research project. It's deployed underground, integrated into the control system, and making automated decisions that operators didn't have before.
Huge credit to Isaac Gallagher and Joshua Mackay who developed and deployed this from concept through to site commissioning.
If your operation has cameras underground and you're not sure they're all actually seeing what they should be — let's talk.
Check out our full blog on our projects page on our website
🌐 https://lnkd.in/gAv-8dGA
#MiningAutomation #MachineVision #AI #UndergroundMining #JMBAutomation #SCADA #LongwallMining #IndustrialAI #JMBGroup
