Fine-grained Air Quality Monitoring Of Smart City Based On UAV Vision

Information technology has become the general trend of social development.The concept of intelligent air quality monitoring system in smart city is to collect and collect data related to air quality at a faster speed,and process it through advanced information technology means to complete air quality monitoring in an instant.Using UAV platform to realize air quality monitoring in smart cities can help people analyze the distribution characteristics of pollutants in time and space,and has the advantages of fast data collection,wide monitoring range and no traffic restrictions.The fine particles that cause urban air pollution have certain spatial variability and dynamic distribution.The traditional restricted air quality monitoring stations have problems such as high cost,poor mobility and difficult maintenance,and can not comprehensively obtain accurate monitoring data.The air quality monitoring is realized by using the mobile intelligent equipment equipped with cameras,which provides a real-time fine-grained air pollution feedback result at a very low cost,and solves the main problems faced by the traditional base stations at present.However,some additional multiple factors that affect air quality cannot be directly reflected in the image,and other methods are needed to fuse the relevant influencing factors.In order to solve the problems of previous air quality monitoring methods,such as inflexibility,lack of mobility and unclear coverage area.This paper designs a fine-grained smart city oriented air particle monitoring system based on UAV vision.By taking images by UAV,the problems such as inflexibility of previous methods and unclear region are solved.In this paper,a user-defined lightweight network module is designed to minimize the amount of parameters and computational overhead without any loss of accuracy.In this module,two branches are made,including standard convolution and deep separable convolution.Finally,the residual mechanism is introduced;Then,using the lightweight neural network framework built with the custom lightweight network module as the core,the atmospheric transmittance features are extracted from the original scene images,which solves the problems of the previous neural network models,such as large computation and memory overhead.Through analysis and verification,multi factor information will closely affect the correlation between PM2.5 and scattering coefficient.In order to supplement the lack of image feature information and improve the monitoring accuracy,this paper designs the relationship between PM2.5 and scattering coefficient as a dynamic mode.A dynamic fitting model is designed to fit the dynamic relationship between scattering coefficient and PM2.5 concentration value by integrating multi factor information,which improves the accuracy limitations of previous image feature-based methods.The smart city air quality monitoring proposed in this study has real-time,accuracy and mobility,and conforms to the concept of intelligent air quality monitoring system in smart cities,and has been verified in experiments.