ONLINE POST CONSUMER PACKAGE IDENTIFICATION BY NIR SPECTROSCOPY COMBINED WITH A FUZZYARTMAP CLASSIFIER IN AN INDUSTRIAL-ENVIRONMENT

An optical setup consisting of a high-throughput near-IR (NIR) spectro graph with an InGaAs-array detector and specially designed collection optics is used to record spectra from post consumer packages located o n an industrial conveyor belt. NIR spectra of packages of different po lymers (PE, PET, PP, PS, and PVC) and of a cardboard/plastic compound are recorded between 900 and 1700 nm with 6.3-ms integration time, The visual distinction between the spectra is demonstrated by aid of prin cipal component analysis (PCA) plots. The spectra are further classifi ed by a FuzzyARTMAP neural network. The method is assessed to be suita ble for on-line identification under industrial conditions.