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 were collected by near-infrared hyperspectral imager in the range of 900 ~ 1 700 nm and corrected by black and white to extract the reflectance spectrum data of the region of interest. The principal component analysis Data denoising dimension. The results show that the first three principal components of the cumulative contribution rate of 98.89%. The load factors of the first three principal components were plotted against the wavelength and six characteristic wavelengths were obtained. Discriminant analysis was performed using reflectance spectral data corresponding to the characteristic wavelengths, and three classification models of plastic identification were established. The test results show that the accuracy of the prediction sample is 95.24%, which shows that the model can accurately and effectively classify PE, PP and PET.