基于近红外光谱结合主成分分析和BP神经网络的常用塑料快速鉴别

发布时间：2019-02-18 20:22

【摘要】：为了实现塑料的分类回收,需要对塑料进行快速准确的鉴别。收集了丙烯腈-丁二烯-苯乙烯(ABS)、聚丙烯(PP)、聚乙烯(PE)、聚对苯二甲酸乙二醇酯(PET)、聚苯乙烯(PS)、聚氯乙烯(PVC)、聚碳酸酯(PC)等7种常用的塑料,利用近红外光谱仪分别测得其反射光谱,应用主成分分析和反向传播(BP)神经网络建立模型进行鉴别。首先利用主成分分析提取光谱的特征信息,前8个主成分的累计贡献率达到94.367%,包含了原始光谱的主要信息,将这8个主成分作为BP神经网络的输入,7种塑料的种类作为输出,建立三层BP神经网络模型。每种塑料各30个样本共210个用来训练神经网络模型,各10个共70个用来预测,预测结果准确率达98.571%,能够有效鉴别常用塑料。
[Abstract]:In order to realize the classification and recovery of plastics, it is necessary to identify the plastics quickly and accurately. Acrylonitrile-butadiene-styrene (ABS),) polypropylene (PP),) polyethylene (PE), polyethylene terephthalate (PET), polystyrene (PS), polyvinyl chloride (PVC),) were collected. Seven kinds of plastics, such as polycarbonate (PC), were measured by near infrared spectroscopy (NIR). Principal component analysis (PCA) and backpropagation (BP) neural network were used to establish models for identification. Firstly, the characteristic information of the spectrum is extracted by principal component analysis. The cumulative contribution rate of the first eight principal components is 94.367, which contains the main information of the original spectrum. The eight principal components are used as the input of the BP neural network. Seven kinds of plastics were used as the output, and a three-layer BP neural network model was established. A total of 210 samples of each plastic were used to train the neural network model and 70 were used to predict each kind of plastics. The accuracy of prediction was 98.571. it can effectively identify common plastics.
【作者单位】：中国计量大学光学与电子科技学院;杭州彩谱科技有限公司;
【分类号】：TQ320.77;TP183

【相似文献】