水色遥感反演中不确定性的概率分析

Because of limits in current technologies, black-box models are prevalent in the research of water components inversing from remote sensing data. However black-box models are not universally applicable due to their poor theoretical bases, so it needs to create a theoretical model based on the radiative transfer equation, for analyzing the remote sensing data. The research generalized and constructed a water color remote sensing theoretical model to inverse water componets concentration, then identified relative parameters and analyzed its influences. At the same time, water quality monitoring experiments synchronized with TM satellite were performed in Macau water area and PoYang lake separately for case studies of water color remote sensing inversion. Research indicates: though water color remote sensing theoretical model is difficult to be applied widely at present, analyses of water color remote sensing influencing factors using theoretical model is very useful to improve inversing arithmetric and can also direct designs of the water color remote sensing sensor. Artificial neural network technology is very suitable to identify complicated model because of its strong capabilities of self-studying, self- adapting and fault containing.

由于各种客观限制，水色遥感反演的效果一般不是很好。本研究总结水色遥感反演模型，标识涉及到的参数和影响因素，并开发概率模型表示参数的不确定性，定量研究这些不确定性在反演模型中的传播，得出最终反演结果的不确定性的概率分布，从而发现那些最需要提高测量精度或频度的参数。本研究对于提高遥感反演的费用效益比有很大的指导意义。