基于FY-3A/MERSI数据和支持向量机的太湖水体悬浮物浓度遥感模型  

Estimation of suspended solids concentration at the Taihu Lake using FY-3A/MERSI data

作　　者：张鸽[1] 黄方[1] 龚绍琦[2] 孙德勇[2] 李云梅[3] 

机构地区：[1]东北师范大学地理科学学院,吉林长春130024 [2]南京信息工程大学地理与遥感学院,江苏南京210044 [3]南京师范大学地理科学学院,江苏南京210046

出　　处：《东北师大学报（自然科学版）》2016年第1期148-153,共6页Journal of Northeast Normal University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(41571405;41271196);国家自然科学青年科学基金资助项目(40801145)

摘　　要：以太湖为研究区,基于2009年4月(春季)和2012年8月(夏季)现场观测的总悬浮物质量浓度数据,以及同步过境的两期FY-3A/MERSI卫星影像,采用支持向量机方法构建了悬浮物浓度遥感模型.结果表明:经粒子群PSO算法优化支持向量机参数,选择径向基函数为核函数,以FY-3A/MERSI各波段遥感反射率及其波段组合,即B565,B650,B685,B765,B865和(B865+B650)/(B650/B865)共6组特征数据作为输入参数,所建立的两个季节的悬浮物质量浓度遥感模型决定系数分别为0.89和0.78,均方误差为0.018 5和0.106 1,为最优的水体悬浮物浓度SVM反演模型.Accurate retrieval of suspended solids concentration of inland water body by remote sensing is one of the important approaches for water quality monitoring.Taking Taihu Lake as the study area,based on the FY-3A/MERSI multi-spectral data and the measured data collected in April 18,2009 and August 4,2012,aprediction model of suspended solids concentration using Support Vector Machine（SVM）is constructed.Particle Swarm Optimizer（PSO）is used to optimize the parameters of SVM model and the Radial Basis Function（RBF）is selected as the kernel function.The spectral bands sensible to suspended solids concentration are determined,and the spectral data of combined bands from FY-3A/MERSI are calculated.The results show that when B565,B650,B685,B765,B865and（B865＋B650）/（B650/B865）spectral data from FY-3A/MERSI images are taken as the input data of the SVM,the prediction model has a better performance with the highest determination coefficient R2（0.89 in spring and 0.78 in summer,respectively）.The mean square error（MSE）of predicted suspended solids concentration is 0.018 5and 0.106 1.It could be regarded as the optimal inversion model suspended matter concentration by SVM using FY-3A/MERSI data.

关 键 词：FY-3A/MERSI数据 悬浮物浓度 遥感 支持向量机 太湖 

分 类 号：X87[环境科学与工程—环境工程]