针对我国近岸高浑浊水体区域MODIS短波红外波段大气校正产品中存在的信号饱和及条带问题,利用神经网络模型,采用准同步的HJ-1A/B卫星CCD影像及实测遥感反射率数据对MODIS/Terra水色遥感大气校正产品进行了质量改进。改进后结果与MODIS/Terra遥感反射率产品相比,平均相对误差为13.3%,信号饱和区域修复结果与实测数据各波段平均相对误差为28.2%。结果表明,该方法在保证结果精度的情况下,能有效地修复MODIS/Terra水色波段因为信号饱和而产生的数据空白区域,同时也能较好地解决MODIS/Terra大气校正产品中的条带问题。
For the highly turbid waters along the China coastal region,MODIS atmospheric correction products derived using shortwave infrared(SWIR) bands always have signal saturation and stripe noise.We demonstrate an improvement method with neural network for MODIS/Terra atmospheric correction products using quasi-synchronization HJ-1A/B satellites CCD images and in-situ data.The average relative error of the improved Rrs and MODIS Rrs is 13.3% and the average relative error of the improved Rrs in signal saturation area and in-situ data is 28.2%.The results show that this method is an effective way to significantly repair the blank area in the MODIS/Terra products and remove the stripe noise with acceptable error.