NIR-SWIR-Hyperspectral-Imaging supported surface analysis for the recovery of waste wood

Cascading of waste wood requires a concept for recovery of solid timbers from deconstruction as a source of clean and reliable secondary feedstock for new wood and wood-based products. An essential requirement for the re-use of wood is a sufficient quality of the near-surface areas that must be free of contaminations like coatings or any wood preservatives. Due to the absence of industrial established automatic testing and sorting methods the possible potential for material re-use of recovered wood in the sense of cascading is not utilized so far. Hyperspectral-Imaging (HSI) is a promising method to improve the situation. In the study on hand results according to detection accuracy and limitations of NIR-SWIR-HSI are presented. As input material solid waste wood (e.g. different kinds of hard wood, soft wood, wood with paint or other coatings, particle boards, and medium density fibreboards) obtained from deconstructions is considered. First, the spectral structures of some different kinds of wood and contamination are examined. Desired are the so-called fingerprints according to significant characteristics in the spectra. The results have been incorporated in a database as training set. For classification tasks some decision trees based on PLS-DA (Partial Least Squares Discriminant Analysis) were exploited. These decision trees are then passed to an industrial NIR-SWIR-Hyperspectral-Imager for generating chemical images of the contaminated wood samples. Results of some sorting experiments are presented. The aim of the tests was to find the limits for sorting throughput and purity. The tests revealed that the spectral differences are mostly large enough for automatic wood classification and sorting operations even at presence of inorganic wood preservatives. In this case the detectability and accuracy of classification depends much on preservative concentrations.