Metal-sorting robot using hyperspecral imaging and perClass
University Liège has been using perClass to build a metal sorting system based on hyperspectral imaging. Classifying metals in hyperspectral images is very challenging task
Today, typical applications of hyperspectral imaging are in agro & food and medical applications. Sorting of scrape metal pieces is a challenging problem due to noise and flat reflection responses of metals. Both the near infrared (VNIR, 400-1000 nm) and short-wave infrared (SWIR 1000-2500 nm) are used. Classification algorithms must be not only accurate but also fast in execution. The ability to directly deploy any trained classifier in a real-time sorting system has been one of the reasons why perClass was selected by the researchers.
Project demo covered by RTC Liège (in French):
YouTube video shows the running sorting system that classifies spectral data using perClass:
This research is described in the recent paper: Pierre Barnabe, Godefroid Dislaire, Sophie Leroy, Eric Pirard: Design and calibration of a two-camera (visible to near-infrared and short-wave infrared) hyperspectral acquisition system for the characterization of metallic alloys from the recycling industry, Journal of Electronic Imaging, Vol.24(6), Nov/Dec 2015