一种基于多元统计分析的土壤含水量高光谱反演模型  

A Hyper Spectral Remote Sensing Model of Soil Moisture Content Based on Multivariate Statistical Analysis

作　　者：贾继堂[1] 程琳琳[1] 余洋[1] 王鹏飞[1] 任俊涛[1] 孟浩灿[1] 

机构地区：[1]中国矿业大学(北京)土地复垦与生态重建研究所,北京100083

出　　处：《测绘科学技术学报》2013年第6期614-618,共5页Journal of Geomatics Science and Technology

摘　　要：为建立方便、快速、大尺度区域土壤含水量估测模型,对陕西省横山县实验区83个土壤样本光谱数据进行研究。对光谱数据进行一阶微分变换处理,提高土壤含水量与变换后光谱数据的相关性,根据相关系数的大小,选取1 412,1 549,1 586,1 842,1 976和2 032 nm五个波段的反射率作为最佳建模反演因子,运用多元统计的原理建立土壤含水量反演模型。实验结果表明,利用因子R1 412,R1 549和R1 842组合建立起的预测方程效果最好,预测方程的相关系数为0.960 1,RMSE(中误差)为1.934 2。这表明建立的土壤含水量反演模型是可行的,模型具有较高的精度。To maintain a convenient, fast, large-scale regional soil moisture estimation model, a research on exper- imental spectral data of 83 samples from experimental zone in Hengshan county, Shanxi Province was done. The spectral data were processed through the first-order differential conversion in order to improve the correlation of the soil moisture with the transformed spectral data. According to the correlation coefficient, the reflectance of five bands, l 412, 1 549, 1 586, 1 842, 1 976 and 2 032 nm, were chosen as the best modeling inversion factor, a soil moisture inversion model was established by using the principle of multivariate statistics. The experimental re- sults show that the predictive equation, established based on the combination of the factor R1 412, R1 549 and R1 842, works best, and the correlation coefficient of the prediction equation is 0.960 1, the RMSE is 1.934 2. So, the established soil moisture inversion model is feasible and has relatively high precision.

关 键 词：高光谱遥感 土壤含水量 相关性 多元统计 反射率 

分 类 号：P237[天文地球—摄影测量与遥感] TP79[天文地球—测绘科学与技术]