引用本文:
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 231次   下载 189 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于光谱指数优选的土壤盐分定量光谱估测
郭鹏1, 李华2, 陈红艳1, 刘亚秋3, 盖岳峰4, 任涛5
1.山东农业大学 资源与环境学院, 山东 泰安 271018;2.山东菏泽水利工程总公司, 山东 菏泽 274000;3.山东凯文科技职业学院, 山东 济南 250200;4.山东颐通土地房地产评估测绘有限公司, 山东 济南 250000;5.山东省泰安市农业局, 山东 泰安 271018
摘要:
[目的]探索基于光谱指数的盐渍土盐分估测的最佳技术路线,为研究区土壤盐分定量、快速遥感监测提供理论基础和技术参考。[方法]以山东省垦利县为研究区,野外采样,获取盐分及其主要离子(Cl-,Na+,Ca2+)含量及高光谱数据;然后采用2种思路:①先选取敏感波段,进而构建常见的5种光谱指数;②先任意两波段组合构建光谱指数,进而筛选敏感光谱指数。最后皆采用随机森林方法(random forest,RF)构建土壤盐分及其主要离子的光谱模型。[结果]基于筛选的敏感亮度指数(1 750,1 620 nm)的RF模型精度最高,作为研究区土壤盐分的最佳估测模型,亮度指数作为最佳光谱指数;思路②明确的特征光谱范围涵盖思路①筛选的敏感波段,更有利于光谱特征分析;思路②建模的结果明显优于思路①;确定最佳技术路线为:任意波段两两组合构建光谱指数后,利用相关分析筛选土壤盐分及其主要离子的敏感光谱指数,进而构建其RF模型。[结论]该技术路线适用于黄河三角洲地区土壤盐渍化信息的有效提取。
关键词:  土壤盐渍化  高光谱遥感  随机森林  黄河三角洲
DOI:10.13961/j.cnki.stbctb.2018.03.031
分类号:
基金项目:国家科技支撑计划“华北平原小麦-玉米轮作区高效施肥技术研究与示范”(2015BAD23B02);山东农业大学“双一流”奖补资金资助(SYL2017XTTD02);山东省重点研发计划盐渍土快速改良与地力培肥产品的研发与应用(2017CXGC0306)
Quantitative Spectral Estimation of Soil Salinity Based on Optimum Spectral Indices
GUO Peng1, LI Hua2, CHEN Hongyan1, LIU Yaqiu3, GAI Yuefeng4, REN Tao5
1.College of Resources and Environment, Shandong Agricultural University, Taian, Shandong 271018, China;2.Shandong Heze Hydraulic Engineering Corporation, Heze, Shandong 274000, China;3.Shandong Kaiwen College of Science & Technology, Jinan, Shandong 250200, China;4.Shandong Yitong Real Estate Appraisal and Mapping Corporation, Limited Liability, Jinan, Shandong 250000, China;5.Taian Bureau of Agriculture, Taian, Shandong 271018, China
Abstract:
[Objective] To explore the best technical route for salt salinity estimation based on spectral indices in order to provide theoretical basis and technical reference for the quantitative calculation and rapid remote sensing monitoring of soil salinity in the study area.[Methods] Taking Kenli County of Shandong Province as the study area, samples were collected in the field, the content of soil salt and its main ions(Cl-,Na+,Ca2+)were measured, and the hyperspectra were obtained. Two different methods were used to select the sensitive spectral indices. The first one was to select the sensitive bands of salt and its major ions and then to build five spectral indices. The second one was to combine any two bands and to construct the five spectral indices, and the sensitive spectral indices were then filtered. The random forest(RF) method was used to build quantitative hyperspectral models of soil salinity and ions contents.[Results] The RF model of brightness spectral indices(1 750, 1 620 nm)exhibited the best precision, thus it was the best estimation model of soil salinity in the study area, and the brightness spectral index was the best spectral index. The characteristic spectral range based on the second method covered the selected sensitive bands based on the first method, thus was more conducive to the spectral characteristics analysis. Meanwhile, the salt prediction model built based on the second method was better than that on the first one. Therefore, the best technical route was to construct the spectral indices by combination of any two bands firstly, then to select the sensitive spectral index of soil salinity and its main ions by correlation analysis, finally to build the RF model.[Conclusion] The technical route is suitable for the extraction of soil salinization information in the Yellow River delta.
Key words:  soil salinization  hyperspectral remote sensing  random forest  Yellow River delta