实验技术与管理2025，Vol.42Issue(6)：9-19,11.DOI:10.16791/j.cnki.sjg.2025.06.002
基于"空天地"同步测量的卫星遥感影像大气校正实验方案设计
Design of atmospheric correction experiment for an Air-Space-Ground synchronous measurement framework of satellite remote sensing in aquatic environments
摘要
Abstract
[Objective]Accurate atmospheric correction is critical for retrieving water quality parameters from satellite remote sensing,as>90%of the radiation signal received by optical satellites originates from atmospheric interference rather than water constituents.Traditional correction methods—such as physics-based radiative transfer models(e.g.,moderate resolution atmospheric radiative transfer code),dark pixel algorithms,and empirical approaches—face notable challenges,including computational complexity,dependence on synchronous ground measurements,and spatiotemporal mismatches between satellite overpasses and ground data collection.This study proposes an innovative air-space-ground synchronous measurement framework to improve atmospheric correction accuracy in coastal waters.By integrating unmanned aerial vehicle(UAV)hyperspectral data,ground-based spectral measurements,and satellite imagery,this study aims to reduce reliance on extensive ground sampling,address low-altitude atmospheric noise,and enhance the reliability of chlorophyll-a(Chl-a)concentration inversion,thereby supporting operational water quality monitoring.[Methods]The experiment was conducted in three phases.First,during Sentinel-2 overpasses,synchronous data acquisition was carried out.UAV hyperspectral imagery(Cubert S185 sensor,450-998 nm,138 bands),ground-measured water surface spectra(TriOS AWRMMS system),and in situ Chl-a concentrations(EXO multi-parameter analyzer)were collected within±1 h in Tangdao Bay,Qingdao.A grid of 29 sampling points with 100 m spacing ensured spatial coverage and spectral consistency.Second,in the air-ground correction phase,UAV imagery was preprocessed to remove spatial noise using LEE filtering and spectral artifacts using Savitzky-Golay smoothing.The matching pixel-by-pixel-empirical line method(MPP-ELM)was employed to align UAV reflectance with ground measurements.Robust regression minimized outliers,and spectral response functions were applied to harmonize UAV and Sentinel-2 bands,reducing mismatches.Third,for air-space correction,an exponential-trigonometric optimized backpropagation(ETO-BP)neural network was designed to map UAV-corrected reflectance to Sentinel-2 multispectral data.The model incorporated scaled conjugate gradient optimization for faster convergence,dropout layers to prevent overfitting,and ensemble learning to enhance robustness.Training used 1 098 data pairs resampled to 10 m resolution for spatial consistency.[Results]The hierarchical air-ground-space correction achieved superior performance.Corrected Sentinel-2 reflectance showed an R2(coefficient of determination)of 0.92 against ground truth,outperforming traditional physics-based methods(R2=0.72),UAV-only correction(R2=0.65),and ground-only approaches(R2=0.89).Chl-a inversion accuracy improved by 42%after correction,with RMSE(root mean square error)decreasing from 8.7 to 5.1 μg/L.The optimal Chl-a model combined a four-band ratio feature and particle backscattering slope,achieving R2=0.88.UAV hyperspectral data addressed the spatial limitations of sparse ground sampling,enabling"face-to-face"correction and reducing false alarms by 35%.Sensitivity analysis indicated that aerosol heterogeneity and residual low-altitude scattering—despite ground-based constraints—remained challenges,emphasizing the need for real-time meteorological integration.[Conclusions]This study presents a robust framework for atmospheric correction by synergizing multiplatform remote sensing data.The key contributions of the proposed framework are as follows.(1)Integration of MPP-ELM with ETO-BP:The hybrid model bridges spatiospectral gaps between ground points,UAV imagery,and satellite pixels,effectively addressing nonlinear atmospheric effects.(2)Operational efficiency:UAVs allow rapid,high-resolution data acquisition,reducing dependence on labor-intensive ground campaigns while maintaining correction accuracy.(3)Enhanced Chl-a retrieval:The model's improved accuracy supports coastal water quality monitoring,particularly in optically complex regions.However,the framework's performance depends on minimal ground data for low-altitude noise suppression and remains sensitive to aerosol variability.Future work should incorporate real-time atmospheric profiles and extend validation to diverse aquatic environments.This approach underscores the potential of multiplatform remote sensing for advancing environmental management and ecological conservation.关键词
大气校正/多源数据/水色遥感/高光谱/BP神经网络Key words
atmospheric correction/multisource data/water color remote sensing/hyperspectral/backpropagation neural network分类
信息技术与安全科学引用本文复制引用
崔建勇,张旺辰,侯舒航,任鹏,刘善伟,宿新元..基于"空天地"同步测量的卫星遥感影像大气校正实验方案设计[J].实验技术与管理,2025,42(6):9-19,11.基金项目
国家自然科学基金重点项目(U1906217) （U1906217）
国家重点研发计划项目(2017YFC1405600) （2017YFC1405600）
山东省自然科学基金面上项目(ZR2023MD115) （ZR2023MD115）
