【作者】 李宁；

【导师】 韩旭； 陶友瑞；

【作者基本信息】 河北工业大学 ， 工程（专业学位）， 2020， 硕士

【摘要】 当前人们不满足于传统的、效率低下的人工分类垃圾的方式,开发新型、智能型垃圾分类装置越来越成为现代垃圾分类的重要任务。本文基于图像信息识别对垃圾分类装置的机械系统、下位机控制系统以及上位机识别程序进行设计,主要内容如下:首先,分析新型垃圾分类装置的发展趋势,提出两种自动分类方案:(1)采用识别垃圾袋颜色的方式对可回收、不可回收、有害等类别的垃圾分类;(2)使用图像处理技术和深度学习技术对可回收类单个垃圾进行分类。根据提出的两种方案,进行垃圾自动分类装置的具体方案设计,包括适应两种分类方案的箱体框架单元、分运机构单元及运输收纳箱单元结构方案设计,和垃圾自动分类装置控制方案设计,从而实现收纳、识别与分类等功能。其次,对两种分类方案的分析基础上,利用工程设计软件对垃圾自动分类装置的整体结构进行设计,建立出装置各机械单元的3D模型,并采用有限元分析软件对关键受力零部件进行静力学分析,验证其可靠性。在对颜色识别自动分类装置控制方案分析的基础上,确定下位机工作流程以及各动作之间的衔接关系,并且对装置下位机的控制程序及电路进行设计。在对图像识别自动分类装置方案分析的基础上,采用Resnet-34残差神经网络及Py Torch框架,设计出单个垃圾图像的识别程序并进行测试,结果表明该程序识别垃圾图像的正确率达到97.3%。最后,研制垃圾自动分类装置样机,进行整机调试。实验结果表明,颜色识别自动分类装置可实现多个垃圾同时投递,且可以采用识别垃圾袋的颜色实现垃圾的自动分类。更多还原

【Abstract】 At present,people are not satisfied with the traditional and inefficient way of artificial waste classification.Developing new and intelligent waste classification device has become an important task of modern waste classification.Based on image information recognition,this paper designs the mechanical system,the lower computer control system and the upper computer recognition program of the garbage classification device.The main contents are as follows:First of all,the development trend of the new garbage classification device is analyzed,and two automatic classification schemes are proposed:(1)the recyclable,non-recyclable and harmful garbage are classified by the way of recognizing the garbage bag color;(2)the recyclable single garbage is classified by using image processing technology and deep learning technology.According to the two schemes proposed,the specific scheme design of the garbage automatic classification device is carried out,including the structural scheme design of the box frame unit,the distribution mechanism unit and the transportation storage box unit adapted to the two classification schemes,and the control scheme design of the garbage automatic classification device,so as to realize the functions of storage,identification and classification.Secondly,based on the analysis of the two classification schemes,the overall structure of the garbage automatic classification device is designed by using the engineering design software,and the 3D model of each mechanical unit of the device is established,and the static analysis of the key stressed parts is carried out by using the finite element analysis software to verify its reliability.Based on the analysis of the control scheme of the color recognition automatic classification device,the work flow of the lower computer and the connection between the actions are determined,and the control program and circuit of the lower computer of the device are designed.Based on the analysis of the scheme of the automatic classification device of image recognition,the recognition program of a single garbage image is designed and tested by using resnet-34 residual neural network and pytorch framework.The result shows that the correct rate of the program is 97.3%.Finally,the prototype of the automatic garbage sorting device is developed and the whole machine is debugged.The experimental results show that the automatic classification device of color recognition can deliver multiple garbage at the same time,and the automatic classification of garbage can be realized by using the color of the garbage bag.更多还原