Quality Analysis And Research Of Satellite Radiation Data

The quality evaluation of remote sensing data plays an important role in the application and development of remote sensing technology,and is the hub of remote sensing instrument development and remote sensing data application.It can not only make a reasonable analysis of the results of previous work,but also provide scientific basis for subsequent satellite launches.Goal clear sky simulation based on ocean satellite observations of the quality of diagnostic analysis is a field application more,because before accurately simulated with 6 sv and time-consuming longer problems,this paper proposes a fast radiative transfer model based on XGBoost regression model,and analyses the simulation value with the model prediction research.The main contents of the paper are as follows:The evaluation of remote sensing data quality plays an important role in the application and development of remote sensing technology and is the hub for the development of remote sensing instruments and the application of remote sensing data.It can not only make a reasonable analysis of the results of previous work,but also provide a scientific basis for the launch of subsequent satellites.Diagnostic analysis of satellite observation quality based on clear sky simulation of ocean surface targets is a field with many applications.Due to the previous accurate simulation with 6SV,there is a problem that it takes a long time,this paper proposes a fast radiation transmission mode based on XGBoost regression model,and analyzes and studies the simulated value and the model predicted value.The main content of the paper is as follows:1、Based on the reflection file of the reflection band observation simulation under the global deep sea clear sky and ocean surface target of Aqua MODIS,establish a data filtering strategy for different wave bands,by aerosol,wind speed,flare angle,solar zenith angle,satellite zenith angle and other data Set the filtering range to filter out local anomalies in satellite data.2、Screen the key impact factors of the deviation,analyze the correlation between the key impact factors and the deviation,and perform feature screening based on the model fusion mechanism of random forest and XGBoost algorithm.Through feature importance calculation,the contribution degree of each feature and the feature ranking result are obtained.After calculation,the key influencing factors selected were flare angle Glint A,column number Col,satellite zenith angle Sen Z,solar zenith angle Sol Z,longitude Lon,latitude lat,aerosol AOD,row number Row,wind speed WS.Then,construct training and test samples,perform various combinations based on the selected key impact factors,and use the polynomial regression model and XGBoost regression model for modeling analysis and prediction,and analyze the deviation distribution characteristics of the predicted value and the simulated value,and the global distribution map.,Dependency analysis of deviation and impact factors,etc.According to MAE,RMSE,R~2 evaluation standards,the model prediction effect is evaluated.3、Using Python as the development language,a satellite radiation data quality analysis system is implemented.The system visually displays the results of satellite data analysis,including satellite data quality control screening module,simulation and observation result mapping module,deviation impact factor analysis module,and remote sensor Sensing space correction modeling module,etc.Through the application of data preprocessing and analysis and modeling algorithms in machine learning,combined with the knowledge of data visualization,the results of various functional modules are displayed to provide information support for the analysis of radiation data quality in the visible light reflection band,and it can also simplify operations and The complex calculation process of researcher data simulation.