Aging civil infrastructure leads to deteriorating assets and the need for regular
manual corrosion inspection. However, current inspection practices are often conducted manually,
which is costly, time-consuming and inaccurate, because the size of corrosion areas is estimated
by hand. In contrast, hyperspectral imaging on Unmanned Aerial Systems (UAS) enables the
automatic remote detection of iron(III)-specific reflectance spectra of steel corrosion products and
its quantitative determination.
In this paper, we present a spectral matching algorithm with α-FeIIIO(OH) as a reference spectrum
and a workflow for quantitative iron(III) detection by hyperspectral image segmentation.
Data is acquired over a length of 270m of a bridge cap under traffic within a flight time of 10 min.
The 272 hyperspectral images are annotated manually with AI assistance and segmented with the
proposed algorithm with an average of 0.677 s per image.
The presented workflow allows the objective and quantitative determination of steel corrosion
areas in compliance with ISO 4628-3 in remote places.