The Research On Automatic Cutting Technology Of Compressor Based On Vision

With the development of society,people pay more and more attention to the construction of ecological civilization.With a large population in China,there is a huge demand for electronic products.The speed of upgrading of electronic products is accelerating,resulting in a huge amount of e-waste,which contains rich renewable resources.In the treatment of e-waste,it is relatively difficult to deal with the refrigerator at present.The reason is that there are many kinds of "compressors" in the refrigerator,and the situation is complex.Therefore,the automatic cutting of compressors has become an urgent problem in the industry.At present,some researchers have applied machine vision technology to the research of compressor automatic cutting,but they have not achieved good results.This paper aims to rely on machine vision technology,design compressor automatic cutting solutions to meet the industrial production conditions,and solve the problem of recycling industry refrigerator compressor processing.First,based on the industry production experience,this paper defines the basic principles that compressor automatic cutting technology needs to meet,and transforms the artificial experience into machine vision task,and defines the machine vision needs to solve the classification of tank body and the key cutting point positioning problem.On this basis,this paper designs a set of automatic cutting solutions for compressors based on machine vision.Secondly,on the basis of the designed solution,this paper completes the research and implementation of the technical points in the scheme.In this paper,a closed acquisition environment is designed to reduce the interference of the environment on the image.In this environment,a large number of compressor image data are collected,and a set of good image preprocessing scheme is developed through experiments,which lays the foundation for the follow-up visual tasks.In this paper,the tank classification task and the cutting key point detection task are completed by using yolov3 network and CPN network,respectively In order to improve the detection effect,this paper optimizes CPN network in three aspects: introducing attention mechanism,modifying refined network loss function,optimizing network structure,and verifies the optimization effect through experiments;after obtaining the image coordinates of key cutting points,this paper uses monocular camera positioning principle to complete the 3D coordinates acquisition of cutting points.So far,this paper has completed the realization of the key technology of the compressor automatic cutting solution based on vision.Finally,this paper uses the above research content to build the system,and carries out 18 consecutive days of cutting experiments in the industrial environment.The success rate of compressor cutting is 94%,and the number of failures caused by visual inspection accounts for about 0.32%.The overall effect is good,which proves that the performance of the research results in this paper is good.To sum up,this paper proposes a set of compressor automatic cutting solution based on machine vision,optimizes the network model used,and verifies the performance of the proposed scheme through experiments.The content of this paper has certain reference significance and industrial value for network optimization and compressor processing.