Research On The Method Of Multi-source Data Fusion In Haze Monitoring

Multi-source data fusion haze monitoring is a large-scale monitoring of haze through multi-source remote sensing technology.Its main purpose is to achieve real-time,rapid,macro and accurate monitoring of haze by utilizing the advantages of remote sensing technology that is not limited by space scope.At present,remote sensing monitoring technology can be used in agriculture,military,Marine,environmental and other fields.In haze monitoring by remote sensing,geometric correction,fusion,aerosol inversion and other operations should be carried out on remote sensing images of the target area to obtain the aerosol thickness value of the study area.Due to the cumbersome process of processing remote sensing images,it is of great significance to accurately and efficiently invert the aerosol thickness of the research area to predict PM2.5.This paper first focuses on the spatio-temporal fusion algorithm of remote sensing images.Multi-source remote sensing images need to be fused,and the types of fusion algorithms vary according to different sensor types.The commonly used spatio-temporal data fusion methods include the traditional spatio-spectral fusion method and the hybrid pixel decomposition algorithm based on linear model.In the research process of this paper,the adaptive fusion algorithm(STARFM)is proposed to improve the algorithm and image preprocessing.The improved image fusion algorithm has a strong feasibility in the fusion of MODIS and gf-1 data.Secondly,the fusion of data inversion of aerosol.When using single MODIS data for aerosol inversion,corresponding remote sensing images with high spatial resolution cannot be obtained.Therefore,the combination of MODIS and gf-1 data for aerosol inversion can better play the advantages of multi-source remote sensing data with high time and high space,and conduct more accurate environmental monitoring in the research area.Finally,after using DBNS model inversion of aerosols,ground of PM2.5 measured values,and at the same time the weather factors,these three kinds of data as input data of the model of PM2.5 projections for the study area,in the concrete study,the study area of shenyang,liaoning province,through the model for average than traditional BP neural network model of prediction error was reduced by 4.89.The research results of this paper show that it is of certain significance to predict haze and control haze pollution.