大气修正是海洋水色遥感中的关键技术之一。近海二类水体大气修正面临两个挑战：浑浊水体造成NIR大气修正波段的离水辐射明显大于零;近海上空存在较强吸收性的气溶胶。本文针对HY-1A CZI,在辐射传输模拟的基础上建立了基于神经网络技术的二类水体大气修正算法,可以由波段1-4的TOA反射率和三个角度反演得到离水反射率、气溶胶光学厚度等参数。利用模拟数据进行了算法的性能评估,并开展了卫星数据处理试验。结果表明,除了在非常浑浊的水体,反演结果基本合理。

Atmospheric correction is one of the key techniques in ocean color remote sensing.Over case 2 waters near shore,atmospheric correction meets two challenges: the nonzero water-leaving radiances in NIR atmospheric correction bands caused by turbid water;aerosol with strong absorption over nearshore regions.In this paper,based on radiative transfer simulations and neutral network techniques,an atmospheric correction algorithm is established for HY-1A CZI in case 2 waters.Water-leaving reflectance and aerosol optical depths can be retrieved from TOA reflectance of band 1 to 4.The algorithm performance is evaluated using simulated data.Satellite data processing experiment shows that the retrieved result seems reasonable except for very turbid waters.