摘要
利用核主成分分析(KPCA)对苹果近红外光谱进行特征提取,不但使得光谱维数大幅降低,而且能有效地提取原始光谱的非线性信息。实验表明,KPCA结合支持向量回归机(SVR)建立的苹果酸度回归模型与PCA-SVR和SVR模型相比,提高了预测精度,缩短了训练时间和预测时间,是一种有效的光谱特征提取方法。
Kernel principle component analysis method, which extracts feature from apples＇ near infrared spectrum, not only drastically reduces dimension but also effectively extracts the nonlinear information from original spectrum. Experiment shows that the regression model for apples＇ acidity, comparing with PCA-SVR model and SVR model, have higher accuracy with less training and prediction time. KPCA is considered to be a effective method of feature extraction from spectrum.
出处
《食品科技》
CAS
北大核心
2012年第3期275-278,共4页
Food Science and Technology
基金
常熟理工学院青年教师科研启动基金项目(QZ1009)
苏州市科技计划项目(SYN201109)
关键词
核主成分分析
近红外光谱
特征提取
kernel principle component analysis
near-infrared spectrum
feature extraction
