宽约束型包装袋垃圾分类模型设计  

Design of the Garbage Classification of the Model Wide-Constrained Packaging Bag

作　　者：周晓莺[1] 余梓唐[1] 全秋燕[1] ZHOU Xiao-ying;YU Zi-tang;QUAN Qiu-yan(Yiwu Industrial and Commercial College,Yiwu 322000,China)

机构地区：[1]义乌工商职业技术学院,浙江义乌322000

出　　处：《塑料科技》2020年第10期93-95,共3页Plastics Science and Technology

基　　金：浙江省教育厅一般科研项目(Y201942569)。

摘　　要：借鉴空域富模型高通滤波器、量化截断机制和卷积神经网络特性,设计宽约束型包装袋垃圾分类模型。研究结果表明:方案八设计的2种量化截断机制捕获到2种约束型残差特征图,有利于特征汇聚;且应用设计的2种子卷积神经网络捕获到多样化包装袋信息,对包装袋垃圾的识别分类准确率为66.2%,高于传统HOG模型约6%。Based on the high-pass filter of the spatial rich model,the quantitative truncation mechanism and the characteristics of the convolutional neural network,a wide-constrained packaging bag garbage classification model is designed.The research results show that:scheme eight uses the two quantitative truncation mechanisms to capture two constrained residual feature maps,which is conducive to feature aggregation,and the two-seed convolutional neural network designed to capture diversified packaging bag information is useful for packaging.The accuracy rate of bag garbage identification and classification is 66.2%,which is about 6%higher than the traditional HOG model.

关 键 词：塑料 包装袋分类 深度学习 子卷积神经网络 量化截断 

分 类 号：TQ619.6[化学工程—精细化工]