Analysis And Implementation Of PM2.5 Concentraion Calculation Module Based On Crowdsourcing Images

PM2.5 particles in the air pollutants are easy to carry bacteria,that can be inhaled into the lungs,the harm to the human body is particularly serious.Officially established air quality monitoring stations are very sparse in many regions because of their high cost,and they cannot effectively monitor the air quality of all locations.With the development of artificial intelligence,machine learning technology is widely used,and historical data is analyzed by designing different algorithm models to infer current or future events,playing a significant role in the fields of image,speech,and recommendation.The development of the Internet of Things has also led to the popularization of crowd-sensing technology and crowd-sourced technologies,and the acquisition of a large amount of diverse data by sensors in the Internet.The work described in this thesis is based on the crowdsourced perception system of the research group.Based on the collected scene data,cell phone models,pose values,weather,seasons,time,and other environmental information,the feature vectors included in the same scene images are excavated.The features has strong correlation with PM2.5 concentration.Based on the characteristics of mining,the PM2.5 concentration calculation model algorithm is studied and designed.A kernel function-based Bayesian parameter estimation method is proposed for different scenes to infer the PM2.5 concentration value.The flow of the algorithm is designed and implemented in a crowdsourced perception platform system that embeds the topic into a crowdsourced platform for crowdsourcing.This makes it possible to establish PM2.5 concentration value inference models for crowded locations and others to achieve fine grained monitoring PM2.5 concentration in spatial and temporalThis thesis first introduces the research background,work objectives and content,and introduces and summarizes related technologies.After that,based on image data,the analysis and design of PM2.5 concentration calculation algorithm was carried out,and describe the methods of data preprocessing,mining features of images that have correlation with PM2.5 concentration.Then training and updating the PM2.5 concentration inference model was obtained based on the characteristics of images in the same scene.And the algorithm is tested and compared to verify the accuracy.Then the requirement description and feasibility analysis are carried out.According to the demand analysis results,the module is outlined and detailedly designed.The realization process of the entire PM2.5 calculation module is described through the description of the introduce of different sub-modules.Finally,the author of this paper tested the designed PM2.5 concentration value calculation module to verify the feasibility of the algorithm and the effectiveness of the system.