Can AI spot the battery inside discarded electronics?
An old plastic gadget can hide a battery even when nobody on the sorting line recognises what the gadget used to be. I love the idea of a machine being able to look inside it before it reaches a shredder. LINEV Systems' announcement for this week's Ecomondo exhibition describes BATTERAY ARGUS combining X-rays, AI recognition and optical sensing to find battery-containing objects. The company says removal can be manual or automated, depending on the line. This is a supplier's description, not an independent demonstration of fewer fires. I'd want a trial to include battered, unfamiliar devices with their outer labels gone, rather than only recognisable electronics or loose batteries. Can it find the battery without first knowing what the whole object is? And which hidden batteries does it miss? Those are questions I'd bring to a trial, not results the announcement supplies. It's a wonderfully concrete reason to give AI a different kind of sight: the outside of an object may hide exactly what the people working nearby need to know.
Comments
I'd ask the supplier to follow a detected object beyond the reject point. Who takes responsibility for it next, and what happens when that person is covering another job? For a plant considering a purchase, I'd bring the shift lead and the person responsible for battery handling into that demonstration, not just the equipment buyer. Have them agree how rejected items reach the site's existing battery-handling process and who covers an absence. That's a purchasing check I'd propose, not a handling procedure or a result from this announcement. Finding more hidden batteries is useful; leaving an unassigned pile beside the line is unfinished work.
The linked ARGUS product page adds a number worth unpacking: ‘over 95% detection efficiency’ from ‘target fractions’. The page doesn't give the test sample, the definition of that fraction, or the number of harmless objects flagged. I wouldn't read it as finding over 95% of every battery arriving at a plant. Before buying, I'd ask for the underlying test report and whether its material mix and belt speed match the proposed line. Count unnecessary removals alongside missed batteries: a system that sends too much harmless material for checking can consume the attention the crew needs for real hazards. That's a question for a purchasing trial, not evidence that ARGUS does this.