【摘要】 高光谱遥感应用于内陆湖泊水质监测具有较好的发展前景,但由于内陆湖泊水体光学环境的时空多变性,如何高效利用水体高光谱特征信息,降低数据冗余度,发展高精度的水质参数反演模型具有重要的意义。针对上述问题,以巢湖为例,将遗传算法和地统计学相结合,利用环境一号(HJ-1A)卫星HSI高光谱遥感数据,建立了基于协同克里格遗传算法的湖泊水质总磷浓度高光谱遥感反演模型。实验结果显示,与传统遗传算法比较,协同克里格遗传算法模拟的ME、RMSE分别提高了128.2%、53%。经总磷实测值和反演值比对,建模和检验的相关系数R2分别为0.85、0.77。反演结果表明:协同克里格遗传算法通过利用克里格插值对传统遗传算法目标函数优化改进,使其具备克里格最佳局部估计能力,能够有效提高反演的精度。更多还原

【Abstract】 Hyperspectral remote sensing is attractive to various applications for monitoring a variety of nutrient parameters of inland lakes.However,for the spatial and temporal variability of optical properties in inland lake,it is of great importance to efficiently use hyperspectral characteristic information of water bodies,reduce data redundancy,and develop high-precision retrieval model of water quality parameters.In the paper,hyperspectral retrieval model of total phosphorus(TP),taking Chao Lake as an example,was established by co-Kriging GA to retrieve TP concentration from HJ-1A satellite HSI hyperspectral remote sensing data.This novel approach was a combination of genetic algorithm and Geostatistics.Experimental results show that ME,RMSE of co-Kriging GA are increased by 128.2%,53% compared with the traditional genetic algorithm.By the fitness of inversed values and measured TP,correlation coefficient R2 of modeling test is 0.85,0.77.Finally,this model is successfully applied to process the images taken by HJ-1A satellite and generates spatial distribution map of TP concentration.The retrieval results show that co-Kriging GA has the best local estimate ability by optimal design to objective function of traditional genetic algorithm,and is validated to improve the retrieval accuracy.更多还原