摘要
实现生活垃圾自动分类是解决城市固体废弃物(municipal solid waste, MSW)问题的有效途径。着眼于近10年基于计算机视觉的垃圾图像识别相关研究,依据垃圾自动分类方法的差异性,将当前现有相关研究分为基于传统机器学习方法和基于深度学习方法。介绍了机器学习方法以及深度学习方法特征提取方式,对比分析了传统机器学习方法和基于深度学习方法的垃圾种类识别的优缺点,着重阐述深度学习方法通用神经网络的应用研究。此外,对当前垃圾图像识别相关研究所用数据集进行了介绍,并对当前垃圾图像识别存在的问题进行了分析与展望。
Realizing the automatic classification of domestic waste is an effective way to solve the increasing problems on municipal solid waste(MSW). The thesis focused on the researches on waste image recognition based on computer vision in the past ten years. According to the differences of automatic waste classification methods, the current existing related research was divided into traditional machine learning methods and deep learning methods. It illustrated the machine learning method and the feature extraction method of the deep learning method, compared and analyzed the advantages and disadvantages of the traditional machine learning method and the waste type recognition based on the deep learning method, focused on the application research of the general neural network of the deep learning method. In addition, the data sets used in the current research on waste image recognition were introduced, and the problem of current waste image recognition were analyzed and prospected finally.
作者
金佩薇
姚燕
梁晓瑜
蔡晋辉
JIN Peiwei;YAO Yan;LIANG Xiaoyu;CAI Jinhui(College of Metrology&Measurement Engineering,China Jiliang University,Hangzhou 310018,China)
出处
《环境工程》
CAS
CSCD
北大核心
2022年第1期196-206,共11页
Environmental Engineering
关键词
图像识别
深度学习
特征提取
卷积神经网络
MSW
image identification
deep learning
feature extraction
convolutional neural network
MSW
