基于深度学习的智能垃圾分类系统设计  

Design of Intelligent Garbage Classification System Based on Deep Learning

作　　者：申立朝 刘清 张玮宸 戴艺 孙慧霞[1] 

机构地区：[1]运城学院,物理与电子工程系,山西 运城

出　　处：《人工智能与机器人研究》2022年第2期104-113,共10页Artificial Intelligence and Robotics Research

摘　　要：目前垃圾分类的前端处理主要依靠人力进行,导致垃圾分类的效率低下且费用高昂。针对这个情况我们设计并开发了一种基于深度学习的智能垃圾分类系统。该系统通过图像识别自动对垃圾分类并投放到相应位置,节省了工人再次分拣的步骤。同时该系统设计了智能识别、控制分类、GPS定位、太阳能充电以及智能烟雾报警等功能,并开发了微信小程序进行远程监测,实时显示每个垃圾桶容纳余量和城市垃圾桶投放量以及投放点。经实验测试,系统功能稳定,垃圾平均识别率可达93%,在解决垃圾分类问题方面具有一定的应用价值。At present, the front-end treatment of waste classification mainly depends on manpower, which leads to low efficiency and high cost of waste classification. In view of this situation, we designed and developed an intelligent waste classification system based on deep learning. The system automatically classifies the garbage and automatically puts it into the corresponding position through image recognition, which saves the steps of workers’ secondary sorting. At the same time, the system designs the functions of intelligent identification, control classification, GPS positioning, solar charging and intelligent smoke alarm, and develops a WeChat applet for remote monitoring to display the capacity allowance of each trash can, the amount and point of urban trash can in real time. The experimental test shows that the system has stable function and the average recognition rate of garbage can reach 93%. It has certain application value in solving the problem of garbage classification.

关 键 词：垃圾分类 深度学习 图像识别 微信小程序 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程]