基于多元统计分析的土壤铬含量高光谱估测模型  

Hyperspectral Estimation Model of Soil Chromium Content Based on Multivariate Statistical Analysis

作　　者：路杰晖 王凤华[2] 刘志文[2] 王德强[2] 李西灿[1] 王旭飞 LU Jie-hui;WANG Feng-hua;LIU Zhi-wen;WANG De-qiang;LI Xi-can;WANG Xu-fei(Geological Surveying and Mapping Institute of Shandong Province,Jinan 250002,China;College of Information Science and Engineering/Shandong Agricultural University,Tai'an 271018,China)

机构地区：[1]山东农业大学信息科学与工程学院,山东泰安271018 [2]山东省地质测绘院,山东济南250002

出　　处：《山东农业大学学报（自然科学版）》2018年第1期144-147,共4页Journal of Shandong Agricultural University：Natural Science Edition

基　　金：山东省地矿局地质科技攻关项目(KY201517)

摘　　要：快速测定土壤重金属的含量,对防治土壤环境污染具有重要意义。本文以山东省烟台市的70个棕壤样本为研究对象,对室外光谱数据进行多种数学变换,根据极大相关性原则从光谱的一阶微分、对数倒数的一阶微分和倒数的一阶微分三种变换中选取5个波段作为反演因子,即:R_(1910.5)(一阶微分)、R_(674.1)(对数倒数的一阶微分)、R_(1609.4)、R_(1231.3)、R_(1127.3)(倒数的一阶微分),然后利用多元统计分析方法和BP神经网络方法分别建立土壤金属铬含量高光谱估测模型。实验结果表明,当利用多元统计分析方法建立估测模型时,14个检验样本的平均相对误差为4.906%,模型的决定系数R^2=0.971,其效果优于BP神经网络模型。研究表明,利用多元统计分析法建立土壤金属铬含量高光谱估测模型是有效的。Rapid determination of heavy metal content in soil is of great significance to the prevention and control of soil environmental pollution.A total of 70 brown soil samples from Yantai City of Shandong Province were as the research object,a series of mathematical transformations were carried out on the outdoor spectral data.According to the principle of great correlation,five bands were selected as the inversion factors from first order differential,first order differential of reciprocal logarithm and first order differential of reciprocal.Namely,R1910.5(first order differential),R674.1(first differential of the reciprocal of a logarithm),R1609.4,R1231.3,R1127.3(first order differential of reciprocal),then the hyperspectral estimation model of soil chromium content was established by multivariate statistical analysis method and BP neural network method.The experimental results show that when using multivariate statistical analysis method to establish estimate model,the average relative error of 14 test samples was 4.906%,the decision of the model coefficient R2=0.971,the effect was better than that of BP neural network model.The study shows that it is effective to establish the hyperspectral estimation model of soil metal chromium content by using multivariate statistical analysis method.

关 键 词：高光谱遥感 估测模型 多元统计分析 土壤铬含量 

分 类 号：X131.3[环境科学与工程—环境科学]