FOD-Detection.com

FODetect becomes an AI product, and the object list gets longer

Xsight has repositioned FODetect around artificial intelligence. The sensing is unchanged — millimetre-wave radar fused with electro-optical imaging at the runway edge, a full runway scan claimed in under sixty seconds with no blind spots — but the claim has moved from finding things to knowing what they are.

The published object list is now broad: broken pavement, loose hardware, building supplies, boulders, sand, luggage fragments, catering equipment and wildlife.

Classification is a real advance and it is worth saying why. A detection system that reports an object requires somebody to drive out and look. A system that reports “metal, 40mm, at threshold plus 1,200 metres” lets an operations centre triage: some objects justify stopping traffic, most do not. Since Changi has just demonstrated that the binding constraint on these systems is nuisance alerts rather than raw sensitivity, anything that helps an operator decide what to ignore is attacking the right problem.

Two things to hold on to, though.

Pavement damage appearing in the object list is the same insight Moog bought a pavement survey company for six months earlier. Both vendors have independently concluded that the surface itself is a debris source worth watching, which is probably the most useful thing to come out of this industry in a decade.

And the sixty-second scan is being led with again. It should not be. Scan time is the number this industry has always preferred to publish because it is easy to measure and always sounds good; probability of detection is the number that determines whether the system reduces risk, and the difference between the two is the entire argument of this site. A classifier makes a detected object more useful. It does nothing for an object the sensor never saw, and no amount of machine learning downstream will recover it.

The question to ask any vendor selling AI FOD detection is not what it can classify. It is what fraction of a specified object set it detects, on a wet runway, at night, and who measured it.

Source (Aviation International News)