提出一种智能方法来分析风云三号A星(FY-3A)中分辨率光谱成像仪(MERSI)的长期响应衰变。该方法通过使用迭代加权多元变化检测(IR-MAD)算法识别同一区域、不同时相卫星图像的不变像元,基于这些不变像元的信号变化评估仪器在该时段的响应衰变。先用IR-MAD方法分析图像场景的不变像元,通过两幅图像不变像元的正交回归获取仪器的相对衰变;再对多个图像对的长时间序列进行分析,采用多项式拟合获得传感器衰变随时间的变化曲线。采用所提方法研究了FY-3A/MERSI北非地区和中国西北部地区的数据,并与其他相关研究得到的仪器衰变结果进行比较。结果显示,所提方法与其他方法获取的仪器衰减规律具有良好的一致性(差异在2%以内),而且从北非地区与中国西北地区得到的结果相吻合(大部分通道的差异在1%以内),验证了本文方法的通用性与可重复性。

We develop an intelligent method to monitor the long-term instrumental response degradation of the Fengyun-3A (FY-3A) satellite medium-resolution spectral imager (MERSI). This method assesses the instrumental response degradation during the interval between the capturing of two images using an iteratively reweighted multivariate alteration detection (IR-MAD) algorithm to statistically select invariant pixels from the different-temporal satellite images, which are obtained from the same geographic region. First, the IR-MAD algorithm is used to analyze the invariant pixels from the image scene; subsequently, the orthogonal regression of invariant pixels from two images is conducted to obtain the relative degradation of the sensor during this interval. Next, all the long-term sequence images are processed in a similar manner and the polynomial fitting is used to obtain the relative degradation curve of the sensor over the entire period of time. Herein, we conduct this procedure using the data of FY-3A/MERSI obtained from north Africa and northwestern China, and compare the obtained results with the instrumental degradation results obtained from other relevant researches. The verification results denote that the proposed method is in good agreement with other methods (the difference is less than 2%). The results obtained in north Africa are consistent with those obtained in northwestern China (the differences are less than 1% in majority of the bands), indicating the universality and reproducibility of the proposed method.