基于特征层融合的PCA聚类分析法在烟叶近红外光谱中的相似度分析  

PCA Clustering Analysis Based on Feature Layer Fusion for Similarity Analysis in Near-infrared Spectra of Tobacco Leaves

作　　者：张延琛 ZHANG Yanchen(Ji'nan Cigarette Factory of China Tobacco Shandong Industry Co.,Ltd.,Ji'nan,Shandong 250100,China)

机构地区：[1]山东中烟工业有限责任公司济南卷烟厂,山东济南250100

出　　处：《农产品加工》2025年第2期57-60,65,共5页Farm Products Processing

摘　　要：为提高近红外光谱在分辨烟叶相似度应用中的准确性,将人脸识别技术、主元分析与聚类分析相结合,提出一种基于特征层融合的PCA聚类分析在烟叶近红外光谱相似度的分析方法。首先,对近红外光谱进行标准化与一阶、二阶导数预处理选择要分析得波数区间段,然后应用人脸识别技术分别对一阶、二阶导数的曲线进行图像分割,分别进行PCA求取特征向量,在特征层进行数据融合,再对融合的数据进行第二次PCA分析,通过设置的阈值选择特征向量与特征空间;最后对特征向量进行聚类分析及烟叶相似度的分辨。结果表明,该方法能够准确地对烟草相似度进行分析。To improve the accuracy of near-infrared spectroscopy in resolving similarity in tobacco leaves,a PCA clustering analysis method based on feature layer fusion was proposed by combining facial recognition technology,principal component analysis,and clustering analysis.Firstly,standardized the near-infrared spectrum and preprocess the first and second derivatives to select the wavenumber interval to be analyzed.Then,apply facial recognition recognition technology to segment the curves of the first and second derivatives separately,perform PCA to obtain feature vectors,perform data fusion at the feature layer,and then perform a second PCA analysis on the fused data to select feature vectors and feature spaces based on the set threshold;Finally,cluster analysis was performed on the feature vectors to distinguish the similarity of tobacco leaves.The results indicated that the method proposed in this article could accurately analyze the similarity of tobacco.

关 键 词：近红外光谱 预处理 PCA 欧式距离 特征层融合 

分 类 号：Q657.3[生物学—生物物理学]