基于Sentinel-2影像东北秋季典型湖泊大气校正方法适用性评价

Evaluation of atmospheric correction processors for Sentinel-2 imagery for typical lakes in Northeast China in autumn

本文利用6S(Second Simulation of a Satellite Signal in the Solar Spectrum)、Acolite DSF(Dark spectrum fitting)、C2RCC(Case 2 Regional Coast Color)、SeaDas(SeaWiFS Data Analysis System)、Sen2Cor(Sentinel 2 Correction)、Polymer(Polynomial based algorithm applied to MERIS)和iCOR(Image correction for atmospheric effects)7种大气校正算法,结合松花湖、月亮泡、小兴凯湖实测遥感反射率数据对“哨兵-2号”(Sentinel-2)数据进行大气校正研究,验证算法性能。整体校正结果显示,相较于实测遥感反射率,上述7种大气校正算法均在可见光波段(400~800 nm)呈现不同程度的低估。除C2RCC算法外,其余6种算法校正后的遥感反射率与实测光谱曲线变化趋势基本吻合,其中Sen2Cor算法与iCOR算法性能最佳,Polymer算法性能最差;在单波段校正精度对比中,Sen2Cor和iCOR算法几乎所有波段的均方根误差和平均绝对百分比误差都低于其余5种算法。Sen2Cor算法在560 nm、665 nm和705 nm处校正精度优于其余6种算法,iCOR算法在443 nm和740 nm处有良好的表现,在490 nm处6S算法校正精度最高,拥有最低的均方根误差(0.0059 sr^(−1))和平均绝对百分比误差(21.40%)。结果表明,这7种大气校正算法均可以在一定程度上去除大气影响,增加影像的可用性,Sen2Cor算法和iCOR算法更适用于本文所研究水体或相似水体。

Quantitative inversion of lake water quality parameters is the most widely used field of lake remote sensing,and the atmospheric correction is one of the key steps to determine the inversion accuracy.In this study,seven atmospheric correction processors,e.g.,6S,Acolite DSF,C2RCC,SeaDas,Sen2Cor,Polymer and iCOR,were used to validate the performances of processors with in situ Remote Sensing Reflectance in the Songhua Lake,the Yueliangpao Lake and the Xiaoxingkai Lake,respectively.From the overall comparison results,these seven atmospheric correction processors had low estimation results comparing with in situ reflectance measurement in visible bands(400-800 nm).Of which,the Sen2Cor and iCOR processors has the best performance,and the Polymer processors has the worst performance,considering the accuracy.All processors performed well and showed a similar tendency with in situ reflectance considering MSI bands,excepted for C2RCC.For the validation performance of single bands,the root mean square error and mean absolute percentage error of Sen2Cor and iCOR processors are lower than the other five processors in almost all bands.Sen2Cor processor worked best at 560 nm,665 nm and 705 nm,and iCOR processor performed is better than other algorithms at 443 nm and 740 nm,6S processor has the highest calibration accuracy at 490 nm,with the lowest root mean square error(0.0059 sr^(−1))and average absolute percentage error(21.40%).Further,our study on the validation of seven atmospheric correction processors indicated that they can remove atmospheric influence and increase the availability of imagery,as well as the Sen2Cor and iCOR preformed best and was more appropriate for studied lakes or lakes with similar aquatic environments.