|
| 拉曼光谱结合深度学习算法的塑料分类的研究 |
| Research on classification of plastics by Raman spectroscopy combined with deep learning algorithm |
| 投稿时间：2022-12-30 |
| DOI：10.3969/j.issn.1005-5630.2023.005.005 |
| 中文关键词: 拉曼光谱 一维卷积神经网络 机器学习 塑料制品 定性分类 |
| 英文关键词:Raman spectroscopy one-dimensional convolutional neural network machine learning plastic products qualitative classification |
| 基金项目:国家自然科学基金 (62003047) |
|
| 摘要点击次数: 2479 |
| 全文下载次数: 1806 |
| 中文摘要: |
| 拉曼光谱法能识别塑料制品光谱特征峰，但操作流程繁琐且准确率有待提升，对此提出了基于一维卷积神经网络 (one-dimensional convolution neural network, 1-D CNN) 的塑料制品分类算法，首先建立以聚乙烯 (polyethylene, PE) 、聚丙烯 (polypropylene, PP) 、聚对苯二甲酸乙二醇酯 (polyethylene terephthalate, PET) 和聚苯乙烯 (polystyrene, PS) 为原材料的40种塑料包装样本数据集；然后设计1-D CNN、K近邻 (KNN) 、决策树 (DT) 和支持向量机 (SVM) 4种算法模型进行训练，并在光谱分类流程、模型准确率和鲁棒性等方面进行对比。实验结果表明，1-D CNN在不经过预处理条件下分类准确率达到98.62%，且在60 dB噪声下仍有96.42%的准确率，优于另外3种传统机器学习算法模型。该结果证实，拉曼光谱融合神经网络的多分类方法可提升塑料制品检测性能。 |
| 英文摘要: |
| Raman spectroscopy can identify the spectral characteristic peaks of plastic products, but the operation process is complicated and the accuracy needs to be improved. Therefore, a classification algorithm for plastic products based on one-dimensional convolution neural network (1-D CNN) is proposed. Firstly, data sets of 40 kinds of plastic packaging samples using polyethylene, polypropylene, polyethylene terephthalate and polystyrene as raw materials were established. Then, four algorithm models including 1-D CNN, KNN, DT and SVM were designed for training, and the spectral classification process, model accuracy and robustness were compared. The experimental results show that the classification accuracy of 1-D CNN can reach 98.62% without pretreatment. And the accuracy rate is 96.42% under 60 dB noise, which is better than the three traditional machine learning algorithm models. The results show that the multi-classification method of Raman spectral fusion neural network can improve the detection performance of plastic products. |
| HTML 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
