基于FV-DCNN的塑料垃圾精细分类模型  

Fine classification model for plastic waste based on FV-DCNN

作　　者：吴谊平 张宇[1] 李鸣[1] WU Yiping;ZHANG Yu;LI Ming(School of Information Engineering,Nanchang University,Nanchang 330000,China)

机构地区：[1]南昌大学信息工程学院,江西南昌330000

出　　处：《传感器与微系统》2021年第6期118-120,共3页Transducer and Microsystem Technologies

基　　金：江西省自然科学基金资助项目(20181BAB211019)。

摘　　要：为实现塑料垃圾的高质量回收利用,提出了一种基于Fisher向量(FV)与深度卷积神经网络(CNN)相结合的塑料垃圾精细分类方法。以塑料垃圾近红外光谱图为原数据,将深度CNN与FV编码相结合建立了塑料垃圾精细分类模型。针对几种典型塑料材料设计了验证实验,实验结果表明:该模型分类精度达到了91.25%,使塑料垃圾精细分类成为可能。In order to realize the high-quality recycling of plastic waste, a fine classification method of plastic waste based on Fisher vector(FV)and deep convolutional neural network is proposed.The method uses the near-infrared spectrum of plastic garbage as the original data, and combines the deep convolutional neural network with Fisher vector coding to establish a fine classification model of plastic garbage.The verification experiment is designed for several typical plastic materials.The experimental results show that the classification accuracy of the model reaches 91.25 %,which makes the fine classification of plastic waste possible.

关 键 词：深度卷积神经网络 Fisher向量 光谱图 塑料垃圾 精细分类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]