期刊
APPLIED SPECTROSCOPY
卷 55, 期 8, 页码 960-966出版社
SAGE PUBLICATIONS INC
DOI: 10.1366/0003702011953108
关键词
pattern recognition; principal component analysis; genetic algorithms; machine learning; wood typing; Raman spectroscopy
Raman spectroscopy and pattern recognition techniques are used to develop a potential method to characterize wood by type. The test data consists of 98 Raman spectra of temperate softwoods and hardwoods, and Brazilian and Honduran tropical woods. A genetic algorithm (GA) is used to extract features (i.e., line intensities at specific wavelengths) characteristic of the Raman profile of each wood-type. The spectral features identified by the pattern recognition GA allow the wood samples to cluster by type in a plot of the two largest principal components of the data. Because principal components maximize variance, the bulk of the information encoded by these spectral features is about differences between wood types. The predictive ability of the descriptors identified by the pattern recognition GA and the principal component map associated with them is validated using an external prediction set consisting of tropical woods and temperate hard and softwoods.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据