Application Of Pattern Recognition Techniques To Identification Of ATR Of Wood

Attenuated total reflection infrared spectroscopy (ATR) is a kind of indestructive inspection technology. It has a wide application in substance determination. In this paper, Chinese cedar, polar, larch, Scots pine, chrysanthemum pear and Africa Pterocarpus samples were measured by using attenuated total reflection infrared spectroscopy. The results showed that most peak locations were similar, that's because the main composition of wood are cellulose, hemicelluloses and lignin. Also different wood have its own characteristic composition. So their ATR spectra have indistinct differentiation. The difference of ATR spectra of these wood samples can be enlarged by first derivative. It means that wood identification become easier. It took a long time to identify different wood species and the result was subjective by comparing ATR spectra.Chinese cedar, polar, larch and Scots pine wood samples were identified by using Fisher discriminatory analysis,Logistic discriminatory analysis,SIMCA method and BP neural network method. Fisher discriminatory analysis is a linear method, whose model recognition rate was 95.35% and validation recognition rate was 93.48%; Logistic discriminatory analysis is suitable for 2 categories. Its recognition rate of Chinese cedar and polar was 90%, but which of 4 kinds of wood was only 72%; SIMCA method is a common method for quality analysis of FTIR. Its model recognition rate and validation recognition rate were 98.84% and 97.83% respectively; BP neural network method is a self-learning method, whose model recognition rate and validation recognition rate are both 100%. In this paper, wood recognition rate by using BP neural network is the highest.ATR with pattern recognition technology can be used to identify wood specie.