Three-band scan detects hidden corrosion above 90% in tests

A painted steel surface can look intact while corrosion develops beneath it—a costly blind spot for merchant ships, coastal infrastructure and maritime facilities. Emma Hernández-Suárez and colleagues have tested an algorithm that spots both surface and subsurface corrosion using three spectral bands and a texture measurement, without heating the sample.

The bands are 518 nm, 655 nm and 838 nm. Combined with texture information, they gave the algorithm accuracy above 90% when identifying corrosion beneath coatings of different types and colors. The research was published in PNAS Nexus.

The approach trims the equipment burden. Multispectral cameras had been used to detect corrosion on the surface, while thermal cameras could detect damage under paint but required a heating element. The new method instead extracts corrosion clues from a small set of light wavelengths and the visual texture of the material.

The researchers trained the algorithm on corrosion samples created in a salt spray chamber. They then tested it on a separate set of samples collected from an industrial environment, where the coatings were unknown. That matters because paint type and color can vary in real inspections. Still, the reported figures come from the research team’s testing; the source reports no independent trial in operational conditions.

And so what, concretely? A free algorithm compatible with commercial multispectral systems mounted on drones could help inspect difficult or hazardous structures while reducing equipment costs and exposure for inspectors. The technology remains at the sample-testing stage, so the next practical question is how reliably it performs across full-scale assets and routine field inspections.