高光谱技术联合归一化光谱指数估算土壤有机质含量

发布时间：2018-03-10 08:49

  本文选题：土壤有机质　切入点：高光谱　出处：《光谱学与光谱分析》2017年11期 　论文类型：期刊论文

【摘要】：随着近地高光谱遥感技术的发展,为快速、有效、非破坏性地获取土壤有机质(SOM)信息提供了可能。土壤高光谱波段数据众多,光谱数据变量之间存在较为严重的多重共线性,影响模型复杂结构,而构建归一化光谱指数(NDSI)可以有效去除冗余信息变量,放大光谱特征信息。以江汉平原公安县为研究区,采集56份耕层土样,在室内获取土壤光谱数据,采用"重铬酸钾-外加热法"测定SOM含量,对实测土壤光谱数据(Raw)进行倒数之对数(LR)、一阶微分(FDR)和连续统去除(CR)三种变换,计算四种变换的NDSI数值,分析SOM与NDSI的二维相关性,并对一维、二维相关系数进行全波段范围内的p=0.001水平上显著性检验,提取敏感波段和敏感光谱指数,结合偏最小二乘回归(PLSR)建立SOM的估算模型,探讨二维光谱指数用于建模的可行性。研究表明,二维相关系数相比一维相关系数有不同程度的提升,以LR最为显著,相关系数数值提升约0.26;基于二维相关性分析提取的敏感光谱指数的PLSR建模效果整体优于一维相关性分析提取的敏感波段,其中,NDSILR-PLSR模型的稳健性最优,验证集R2为0.82,模型验证RPD值为2.46,模型稳定可靠,可以满足SOM的精确监测需要,适合推广到区域范围内低分辨率的航空航天遥感(如ASTER,Landsat TM等),应用潜力较大。
[Abstract]:With the development of near-ground hyperspectral remote sensing technology, it is possible to obtain soil organic matter SOM information quickly, effectively and nondestructive. The complex structure of the model is affected, and the normalized spectral index (NDSI) can be used to remove redundant information variables and enlarge spectral characteristic information effectively. Taking Gongan County in Jianghan Plain as the study area, 56 soil samples were collected and soil spectral data were obtained indoors. The SOM content was determined by "potassium dichromate plus calorimetry". The reciprocal logarithmic LRN, first-order differential SOM and continuum removal NDSI were used to calculate the NDSI values of the four transformations, and the two-dimensional correlation between SOM and NDSI was analyzed. The correlation coefficients of one and two dimensions were tested at the level of pn0. 001 in the whole band range. The sensitive band and sensitive spectral index were extracted, and the estimation model of SOM was established by combining with partial least squares regression. The feasibility of using two-dimensional spectral index in modeling is discussed. The results show that the two-dimensional correlation coefficient has different degrees of improvement compared with one-dimensional correlation coefficient, LR is the most significant. The correlation coefficient was increased by 0.26, and the PLSR model based on two-dimensional correlation analysis was better than that of one-dimensional correlation analysis, in which the robustness of NDSILR-PLSR model was optimal. The verification set R2 is 0.82and the model verification RPD value is 2.46. the model is stable and reliable, which can meet the needs of accurate monitoring of SOM. It is suitable for the application of low-resolution aerospace remote sensing (such as ASTER-Landsat TM, etc.) in the regional range, and has great application potential.
【作者单位】： 华中师范大学地理过程分析与模拟湖北省重点实验室;华中师范大学城市与环境科学学院;
【基金】：国家自然科学基金项目(41401232,41271534) 中央高校基本科研业务费专项资金项目(CCNU15A05006,CCNU15A05004)资助
【分类号】：S153.621;TP79
，

本文编号：1592613