Design Of Intelligent Garbage Classification System Based On NB-IoT

With the continuous improvement of living standards,China has gradually entered the era of garbage classification.In recent years,the government has made important instructions on garbage sorting and has called on people to cultivate the habits of garbage sorting.The treatment of recyclable waste is a significant part of the waste classification work,that is,the recyclable waste is identified and classified for reuse.Among them,there are many problems in the process of recyclable waste classification,such as heavy workload,poor working environment,low classification accuracy,and inability to feedback information in a promptly manner.It cannot meet the domestic requirements of the immediateness and accuracy of the classification of recyclable waste in the environmental protection at this stage.In recent years,the technique of IoT(Internet of Things)and machine vision develops continuously.The technique of IoT and machine vision has been universally applied in various industries.However,the lack of suitable recyclable garbage classification system which is restricted by various development factors in environment protection.It is evitable to develop an intelligent garbage classification system combining the two of the technique.Developing an intelligent garbage classification system is of vital important.Firstly,this article introduces the overall design scheme and the composition of system.The system consists of main control board,object recognition unit,and the upper computer on cloud server.Machine vision is used to input the image of classify object,the classify algorithms are used to extract and classify feature of object.The classification result and environment parameters are transmitted through NB-IoT.The data is saved by cloud server for subsequent analysis.In the target recognition unit,a large amount of datasets are used to train four different neural network models due to the complexity of object composition in garbage classification.During the design process,the neural network models under PyTorch framework are optimized and improved.In the recognition process,the classification results are weighted according to the recognition accuracy corresponding to the four neural networks in the training process.The main controller part is designed based on STM32F103ZET6 and a series of sensors to cooperate to detect flammable,explosive and toxic gas in the system and collect overflow status.The communication module is BC35-G which transmits data as NB-IoT module.It can realize the design requirements of sending environment parameters and low power consumption.The enhanced signal penetration makes the communication process more stable,more accurate and more reliable.A guidance system is designed on the main control board,and users can view related information through the LCD screen operation.The cloud server is partially deployed on Tencent Cloud Server.Store and analyze the received information according to the requirements of the design plan.This paper introduces the design scheme of improving the multi-concurrency ability and ensuring high reliability in the process of information transmission.And accomplish online test.Finally,the entire system is validated in a laboratory environment.Experiments on system stability are performed through target recognition,harmful gas detection,communication stability,and many other aspects.