基于高光谱数据和SVM方法的土壤盐渍度反演  

Soil Salinity Inversion Based on Hyperspectral Data and SVM Method

作　　者：池涛[1] 曹广溥 李丙春 孜克尔.阿不都热合曼 王文龙 CHI Tao;CAO Guang-pu;LI Bing-chun;Zikl Abdurahman;WANG Wen-long(College of Information Technology/Shanghai Ocean University,Shanghai 201306,China;College of Computer Science and Technology/Kashgar University,Kashgar 844006,China)

机构地区：[1]上海海洋大学信息学院,上海201306 [2]喀什大学计算机科学与技术学院,新疆喀什844006

出　　处：《山东农业大学学报（自然科学版）》2018年第4期585-590,共6页Journal of Shandong Agricultural University：Natural Science Edition

基　　金：国家自然科学基金(61561027);上海市自然科学基金(16ZR1415100);上海市科技创新行动计划项目:小龙虾生态化智能化设施养殖关键技术研究与应用(16391902900)

摘　　要：在土壤可见光~近红外波段光谱(350~2500 nm)内包含了大量可以表征土壤盐度的相关性信息,通过土壤盐渍化检测反演模型的建立,可以定性定量分析土壤盐渍度。本文探究了信息不完备情况下,机器学习算法在土壤盐渍化检测模型反演中的适用性,分别使用了多元线性回归模型、BP神经网络模型和SVM模型来反演土壤盐渍化检测。结果表明,SVM模型预测精度达到99.41%,是三种模型中精度最高的模型,适用于分析土壤复杂光谱信息和高精度土壤盐渍化反演。There is a large amount of correlation information that can characterize soil salinity in the visible light^near infrared spectrum of soil(350~2500 nm).The inversion model of soil salinization detection can be used to analyze soil salinity qualitatively and quantitatively.In this paper,the applicability of machine learning algorithm in the inversion of soil salinization detection model was explored under the condition of incomplete information.Multivariate linear regression model,BP neural network model and SVM model were used to inverse soil salinization detection.The result showed the prediction accuracy of SVM model was 99.41%,which was the most accurate of the three models.It is suitable for analyzing soil complex spectral information and high precision soil salinization inversion.

关 键 词：光谱数据 支持向量机 土壤 盐渍度 反演 

分 类 号：S151.9[农业科学—土壤学]