基于高光谱数据的农田土壤养分含量估测模型研究  

Research on Soil Nutrient Content Estimated Model by Hyperspectral Remote Sensing Data

作　　者：祁亚琴[1,2] 张显峰[1] 张立福 吕新[2] 张泽[2] 陈剑[2] 李新伟[2] 王飞[2] 彭奎 QI Ya-qin;ZHANG Xian-feng;ZHANG Li-fu;LV Xin;ZHANG Ze;CHEN Jian;LI Xin-wei;WANG Fei;PENG Kui(School of Earth and Space Sciences,Peking University/Institute of Remote Sensing,Beijing 100871, China;Key Laboratory of Oasis Eco-agriculture of Xinjiang Production and Construction Corps,College of Agronomy,Shihezi University,Shihezi Xinjiang 832003,China;Institute of Remote Sensing and Digital Earth,Chinese Academy of Sciences,Beijing 100094,China)

机构地区：[1]北京大学地球与空间科学学院/遥感研究所生态遥感实验室,北京100871 [2]石河子大学/新疆兵团绿洲生态农业重点实验室,新疆石河子832003 [3]中国科学院遥感与数字地球研究所高光谱研究室,北京100094

出　　处：《新疆农业科学》2018年第3期490-495,共6页Xinjiang Agricultural Sciences

基　　金：国家自然科学基金项目"基于高光谱遥感数据的土壤主要养分含量信息的获取研究"(61465011);国家留学基金(201405215035);石河子大学青年教师与对口支援高校名师"结对子"培养(SDJDZ201509);石河子市科技攻关"利用3S技术快速监测农田土壤信息研究与示范"(2013GY01)项目~~

摘　　要：【目的】快速、实时、准确、无损地获取农田土壤主要养分(全氮TN、全磷TP、全钾TK)含量的信息。【方法】运用各种土壤反射率的光谱特征分析技术,提取其最具代表性的敏感波段位置,建立土壤养分含量反演模型。【结果】建立估算模型中,预测TN含量以指数函数模型(Y_(TN)=0.000 5e^(4.700 3x NDI))为最佳;预测TP含量以一元三次函数模型(Y_(TP)=802.27x_(NDI)~3-412.32 x_(NDI)~2+72.357 x_(NDI)-3.318 9)为最佳;预测TK含量以一元三次函数模型(Y_(TK)=80 189 x_(NDI)~3-11 471 x_(NDI)~2+490.57 x_(NDI)+13.879)为最佳模型。【结论】通过模型精度评价和田间反复验证,基于归一化光谱指数NDI建立的高光谱遥感定量模型,能较好的反演土壤TN、TP、TK含量,达到良好的预测效果。【Objective】To obtain the real-time,non-contact and non-destructive information of main nutrients(TN,TP and TK)of farmland soil quickly.【Method】Soil nutrient content model was established by spectral feature analysis techniques and sensitive wavebands.【Result】NDI-based estimates of the soil nutrient content prediction model with exponential function model(Y TN=0.000,5 e 4.700,3 xNDI)was the best to estimate TN;a cubic function model(Y TP=802.27 x NDI 3-412.32 x NDI 2+72.357 x NDI-3.318,9)was the best to estimate TP;a cubic function model(Y TK=80,189 x NDI 3-11,471 x NDI 2+490.57 x NDI+13.879)was the best to estimate TK content.【Conclusion】The results showed that the hyperspectral remote sensing quantitative model based on normalized spectral index(NDI)can be used to invert the content of TN,TP and TK in soil and achieve a good prediction effect through the precision evaluation of the model and repeated verification in fields.

关 键 词：高光谱遥感 土壤 养分含量 预测模型 

分 类 号：S153[农业科学—土壤学]