The Design And Implementation Of PM2.5 Monitoring System Based On Crowd Sensing

Haze threats in a more broader and frequent way,which has exposed people's health and normal life to tremendous danger.In fact,it is average density of fine Particulate Matter 2.5(PM2.5)that is to blame and it gains ongoing concentration.The thesis aims to propose a machine learning based model to forecast the density of PM 2.5,develop a measuring system by adopting the model,which measures the related data around by uploading a picture in a quick and precise way with internet access.Further more,the massive training samples uploaded can be assembled effectively aiming to improve the accuracy under the theory of Crowd Sensing.The thesis makes the literature review of two mature machine learning Algorithms,i.e.Convolutional Neural Network(CNN)and Recurrent Neural Network(RNN)respectively and Algorithm trainings are conducted based on Tensorflow,Google machine learning platform.Further more,big data platforms Hadoop and Mongodb are mentioned due to their massive data support for the training,depend on which a set of big data storage cluster is set up.The realization process is further revealed,including Allibaba's open source PRC framework,Dubbo,aiming at enhancing server's process ability under high concurrency and the clients of Android and IOS.Finally,the functional test,predictive accuracy test and performance test under high concurrency are made on the system,which testifies that the system constructed in this thesis paper has a significant promotion on accuracy and portability compared with other measuring ways of the density of PM 2.5.