Non-ferrous scrap metals classification by hyperspectral and multi-energy X-ray transmission imaging

Hyperspectral imaging in the visible and near-infrared (VNIR) has nevertheless proven to be an efficient technique to identify aluminum, zinc, copper, brass alloys and stainless steel (e.g., Barnabé et al., 2015, Kutila et al., 2005). The efficiency of hyperspectral classification however depends on the surface conditions of the alloy fragments. Coating or dirt can indeed impede the performances of this technique. Multi-energy X-ray transmission (XRT) measurements through the alloy fragments offers another non-destructive way to probe their composition based on the relationship between the atomic number of the absorbing element and the shape of its attenuation curve (Hubbell & Seltzer, 2004). XRT characterization is advantageously insen-sitive to the surface conditions of the fragments.
In this paper, we investigate the added value to non-ferrous alloy classification brought by the combination of VNIR and XRT measurements in a limited sample of metal waste. It should be emphasized that this study focuses on the step of the sorting process during which metallic fragments are characterized and categorized. After a brief presentation of the alloys of interest, equipment used, and data collected, the results of characterization based only on VNIR hyperspectral imaging are compared with the characterization results based on both VNIR and XRT measurements.