Foreign object detection
Bolts, teeth, wire, wood, plastic and rubber — including objects a metal detector cannot see.
The crusher is protected by detecting foreign objects. Tramp or polluted feed is redirected fast and precisely, which reduces downtime from tramp incidents and keeps reject low. Tramp has to be visible to the camera — if it is not in view, it cannot be detected.
A bolt on a belt of crushed aggregate, found in time to be rejected before the crusher.
Partly buried parts, with only a fraction of the object above the fines.
Small parts in dust, at the size where a metal detector stops being selective.
Top size detection
Oversize feed flagged before it reaches the crusher, and an alarm when the crusher clogs.
Coarse-crusher downtime is an unnecessary production loss, and simple top size and clog detection prevents it. We train and develop a model for your specific crusher inlet that flags feed that will not pass and alarms when the crusher clogs, so you get more performance and production out of it.
Top size detection is done by segmentation of the feed. Size as well as shape factor can be detected. Large or oversize rocks can be accurately detected using this method. Odd size rocks that might pose a problem can also be picked out by recognising long shapes.
Ore monitoring
Wetter, drier, finer, coarser, a different orebody — tracked over time so downstream can be adjusted in advance.
Wet feed, different lighting
bone dry feed, different lighting
Feed monitoring is crucial for high recovery and low energy consumption of crushing and chemical processes. Blasting, weather and natural variation all change the ore, and different feed can call for different operating settings — but predicting that change in advance is often very difficult. AI vision can cheaply add information to your data processing stream, it can estimate size ranges, top size and inform you about general deviation in color, size composition on a detail scale that was not possible before.
Start today with acquiring your ore image database to build the dataset for tomorrow. We advise on data acquisition and work with your operators to train that process model.
Discuss an application
Reach out to us and send over your question or footage!