Hyperspectral imaging in qualitative and quantitative assessment of biocide treated timber

A hyperspectral camera operating in the shortwave infrared light range between 1000 – 2500 nm was used to collect spectral profiles of timber treated with different biocides and varying concentrations of a copper-ethanolamine based wood preservative. The spectra were analysed using statistical methods like principal component analysis, linear discriminant analysis, and partial least squares to assess the differences between the spectral profiles of different biocides used in timber treatment and different concentrations of copper-ethanolamine based preservatives. We were able to differentiate some biocides, like CCA, CCB, copper-ethanolamine based on the shortrange infrared spectra but were less successful with quaternary ammonium compound-based biocides, but all samples were significantly different from the untreated timber controls. Additionally, we were able to observe a linear relationship between the shortwave infrared spectra in the partial least squares analysis. Finally, we constructed two basic K-nearest neighbour models for the classification of the used biocides and biocide concentrations. Qualitatively, these were used to some success as we were able to train the models to differentiate different types of biocides semi-effectively. Quantitative determination was also successfully applied even though several classifications were made as false positives.