Study On Simulation Of Images In Near Ultraviolet Channel

The near-ultraviolet spectral range has important applications in fields such as global aurora detection,marine oil spills,and atmospheric fluorescence.The surface reflection characteristics of this range are significant background data for research.However,the existing satellite data resources are relatively scarce and can hardly meet the application demands.In recent years,with the improvement of computer performance and the development of remote sensing technology,remote sensing image simulation has become an important research method,providing more accurate and efficient approaches for studies in various domains.Remote sensing image simulation technology can generate high-quality and realistic remote sensing images by simulating different optical transmission processes of solar radiation energy in the atmospheric and surface transmission processes,such as atmospheric scattering,surface reflection,cloud cover occlusion,etc.,so as to improve the accuracy and credibility of sensor performance evaluation and algorithm verification.The near-ultraviolet entrance pupil radiance image simulation is an effective tool for near-ultraviolet quantitative remote sensing research,which can provide data basis for near-ultraviolet atmospheric correction,surface reflectance inversion and pre-validation,as well as a reference for sensor setup and optimization.This article focuses on the near-ultraviolet range of 350-400 nm,with center wavelengths at 355 nm,365 nm,375 nm,385 nm,and 395 nm（S1-S5 channels）.The surface reflectance images of S1-S5 channels were simulated by analyzing their atmospheric transmission properties,and the entrance pupil radiance images of the corresponding channels were simulated on this basis.The main conclusions are as follows:1.Based on the atmospheric radiative transfer model MODTRAN,we analyze the transmittance of 300-400 nm in the near ultraviolet at different altitudes,atmospheric models,aerosol types,visibility and observation zenith angles;we simulate and analyze the effect of background radiation in this spectral range by different atmospheric models,aerosol types,visibility and surface types.The results show that the 350-400 nm spectral range has high transmittance and stable background radiation values under different atmospheric conditions and observation geometry,and is the visible spectral range for the surface and the sensor,which can be selected as the simulated spectral range for near-ultraviolet images.2.Considering the spectral range settings of the target near-ultraviolet simulation channels,Sentinel-2 MSI multispectral data from channels 2,3,and 4 were selected as the data source.Typical feature spectral data such as vegetation,water bodies and soil were acquired based on the USGS feature spectral database,and the feature spectral data were equivalently calculated to the corresponding data sources and target channels to be simulated.Correlation analysis was conducted between the data sources and the channels to be simulated.The correlation coefficients between Sentinel-2 MSI channels 2,3,and 4 and the channels to be simulated were all greater than 0.88,indicating that the simulation of near-ultraviolet surface reflectance data could be carried out based on this data source.The near-ultraviolet channel surface reflectance regression models were constructed using RR,SVR-LR,SVR-POLY2,SVR-POLY3,SVR-RBF and XGBoost machine learning algorithms based on a typical feature spectral dataset after equivalent calculations.The accuracy results showed that the XGBoost model outperformed the other models,with model coefficients of determination（R²）above 0.910 for all channels,root mean square errors（RMSE）below 0.076,and mean absolute percentage errors（MAPE）below 20%overall.Additionally,the standard deviations of the three accuracy metrics for different sample categories were within a range of 0.021,indicating high accuracy and good robustness of the model.Finally,based on the Sentinel-2 MSI channels 2,3,and 4 image data,the XGBoost model with the best performance was used to generate the surface reflectance simulations at 355 nm,365 nm,375 nm,385 nm,and 395 nm,and the images effectively captured the spectral characteristics of different land cover types.3.A rapid simulation study was conducted for the entrance pupil radiance images of the surface reflectance simulation images in the S1-S5 channels.Based on the imaging conditions of the data source and the atmospheric radiative transfer analytical model,atmospheric parameters were obtained using MODTRAN to account for atmospheric effects.Based on the simulated surface reflectance data,entrance pupil radiance simulation images were generated for the data source area.The results show that the same atmospheric and observation geometric conditions can be used for entrance radiance image simulation in this region.At the same time,by comparing the true values calculated by MODTRAN with the simulated values calculated by the atmospheric radiation transmission analysis model,it was found that the S1-S5 channels all achieved small errors,with error values less than 7.5×10^-5,and the simulation accuracy was higher in the S4 and S5 channels,with error values around 5×10^-5.