基于近红外光谱建立PE、PP和PET的识别分类模型  

Establishment of identification and classification model of PE,PP and PET based on near infrared spectroscopy

作　　者：张毅民[1] 王鹏[1] 白家瑞[1] 马冬雅[1] 

机构地区：[1]天津大学化工学院,教育部绿色合成与转化重点实验室,天津300072

出　　处：《现代化工》2016年第3期182-186,共5页Modern Chemical Industry

基　　金：国家“863”计划资助项目(2012AA063007)

摘　　要：利用近红外高光谱成像仪在900~1 700 nm的范围采集PE、PP和PET样本的高光谱图像,并进行黑白校正,提取感兴趣区域的反射率光谱数据;利用主成分分析法对提取的数据去噪降维。结果表明,前3个主成分的累计贡献率达98.89%。把前3个主成分的载荷系数对波长作图,得到了6个特征波长;利用特征波长对应的反射率光谱数据进行判别分析,并建立了3种塑料的识别分类模型;用预测样本对模型进行检验,结果显示,预测样本的识别准确率为95.24%,表明该模型可准确有效地对PE、PP和PET进行识别分类。The hyperspectral images of PE,PP and PET samples are acquired using near-infrared hyperspectral imager in the range of 900- 1 700 nm. Spectral reflectance data in the region of interest are extracted from the hyperspectral images after black and white correction. The denoising and dimensionality reduction of the extracted data is performed using principal component analysis. The results indicate that cumulative contribution rate of the first three principal components can reach up to 98. 89%. Six characteristic wavelengths are selected according to the graph of the loadings of the first three principal components versus wavelength. Identification and classification model of these three kinds of plastic is set up on the basis of the discriminant analysis results of the spectral reflectance data which corresponds to the characteristic wavelengths. Finally,the model is validated by predicting samples,and results show forecast accuracy rate of the predicting samples is 95. 24%. It turns out that this model can identify PE,PP and PET accurately and effectively.

关 键 词：近红外高光谱 塑料识别分类 特征波长 判别分析 识别模型 

分 类 号：O657.3[理学—分析化学]