摘要
为研究宁夏回族自治区平罗县土壤盐渍化问题,运用Matlab R2016a软件对实测高光谱数据进行小波变换去噪、数值变换,以400~1 800 nm波段的反射率数据作为特征值,建立基于异质支持向量机(SVM)算法的土壤盐渍化分类模型,分析土壤盐渍度。结果表明:最优小波基为db3,最优分解层数为3层,最优阈值为H阈值时,利用最优参数的小波变换可以有效地消除噪声;选择波段800~848 nm,核函数为Polynomial的核函数,参数值c=2.828 4,g=0.933 0,异质SVM模型二分类时准确率最高可达85.0%,四分类时准确率达70%。
In order to study the soil salinization in Pingluo County,Ningxia Hui Autonomous Region,the measured hyperspectral data is subjected to wavelet transform denoising and numerical transform by the Matlab R2016a software.Taking the reflectance data of the band from 400 nm to 1800 nm as the characteristic value,a soil salinization classification model based on heterogeneous support vector machine(SVM)algorithm is established to analyze the soil salinization degree.The results show that,when the optimal wavelet basis is db3,the optimal decomposition level number is 3 and the optimal threshold value is H,the noise can be effectively eliminated by the wavelet transform with the optimal parameters,when the waveband is within 800~848 nm,the Polynomial kernel function is taken as the kernel function and the parameter values c=2.8284 and g=0.9330,the accuracy rate of heterogeneous SVM model can reach 85.0%in dichotomy classification and 70%in quartering classification.
作者
高曦文
贾科利
毛鸿欣
张俊华
GAO Xiwen;JIA Keli;MAO Hongxin;ZHANG Junhua(College of Resources and Environmental Science,Ningxia University,Yinchuan 750001,China;Institute of Environmental Engineering,Ningxia University,Yinchuan 750001,China)
出处
《现代电子技术》
2021年第3期155-161,共7页
Modern Electronics Technique
基金
国家自然科学基金项目(41561078)
宁夏自然科学基金(2018AAC03007)。
关键词
小波分析
支持向量机
高光谱
数值变换
定量分类
土壤盐渍化
wavelet analysis
SVM
hyperspectral
numerical transformation
quantitative classification
soil salinization
