Artificial neural network and SIMCA classification in some wood discrimination based on near-infrared spectra

Abstract

Species recognition and identification are very important in the wood industry. The identification of some tree species is complex when only the wood is available, often requiring multiple characterization techniques. In some instances, near-infrared spectroscopy can provide a method for the identification of wood species. However, as the amount of data acquired by near-infrared spectrometers are large, there is a need to use mathematical and computational tools to treat and analyze the data. This paper reports the results of testing an artificial neural network in comparison with SIMCA classification to identify some Brazilian wood species based on near-infrared spectra. The neural network developed did not result in any identification error for a margin of ±2% with the use of a spectral range from 4000 to 10,000 cm−1, while SIMCA produced more than 60% identification error with raw spectral data. The artificial neural network was more efficient than the SIMCA classification and has good potential to be applied for species discrimination.