Merchant ships, coastal infrastructure, and maritime facilities are often built with iron and steel, which despite their strength, can corrode over time, eventually to the point of failure. With lives and goods on the line, monitoring is essential, but it can be time consuming and inexact when done via visual inspection. Drones equipped with optical systems can help, but multispectral cameras have only been used to spot surface corrosion, and thermal cameras that can see corrosion under paint require a heating element. Emma Hernández-Suárez and colleagues show that information from just three spectral bands (518 nm, 655 nm, and 838 nm), combined with a texture feature, can detect surface and subsurface corrosion without heating the sample. The authors’ generalizable corrosion detection algorithm was built on data from corrosion samples created in a salt spray chamber and tested on a separate set of samples with unknown coatings collected from an industrial environment. The algorithm consistently achieves accuracies above 90% when spotting surface and subsurface corrosion across coatings of different types and colors. According to the authors, the algorithm, which is free to use and compatible with commercial multispectral systems mounted on drones, can help reduce costs and increase safety.
PNAS Nexus
From hyperspectral to multispectral: Robust corrosion detection in iron and steel using minimal bands
25-Aug-2026