DOMINANT TECHNOLOGIES IN “INDUSTRY 4.0”
Wood defects classification based on hyperspectral imaging
Wood processing requires reliable automated inspection and classification of natural defects. Conventional vision methods are often inadequate for distinguishing visually similar defects, particularly bark and resin pockets. This study compares hyperspectral methods for wood-defect classification. Pine samples were imaged in the VNIR (400–1000 nm) and SWIR (1000–2500 nm) ranges. Images were manually labeled for earlywood, latewood, sound and black knots, resin and bark pockets, and pith. Classification workflows combined raw spectra, PCA, or wavelet-based dimensionality reduction with Pearson correlation, k-NN, and Random Forest classifiers. The results enable comparison of the suitability of individual spectral ranges, feature-reduction methods, and classifiers for automated defect identification