摘要
为了预测土壤pH值和全钾含量,采集111个土壤样本的高光谱信息,采用小波变换后去包络的方法对原始光谱信息进行预处理,利用相关性分析法选择土壤光谱特征波段,进行偏最小二乘法回归(PLSR)、主成分分析回归(PCR)、支持向量回归(SVR)3种方法的土壤pH值和全钾含量高光谱反演精度的比较研究。结果显示,在水稻土和紫色土全钾含量和水稻土pH值的反演中,SVR方法都取得了比PLSR方法和PCR方法更好的反演效果。在紫色土pH值反演中,PLSR方法和PCR方法反演效果均优于SVR方法。比较不同类型土壤和不同土壤参数的反演效果发现,水稻土土壤pH值和全钾含量的反演效果均强于紫色土,全钾含量的反演效果优于pH值。本研究结果说明高光谱快速反演土壤pH值和全钾含量具有可行性。
In order to predict soil pH value and total potassium content,hyperspectral information of 111 soil samples was collected.The original spectral information was pretreated by continuum removal after wavelet transform.The characteristic bands of soil spectra were selected by correlation analysis method.A comparative study on hyperspectral inversion accuracy of soil pH and total potassium content was carried out by using partial least squares regression(PLSR),principal component analysis regression(PCR)and support vector regression(SVR).SVR method was better than PLSR method in the inversion of total potassium content of paddy soil and purple soil and pH value of paddy soil and PCR method in the inversion of total potassium content of paddy soil and purple soil and pH value of paddy soil.In the inversion of pH value of purple soil,the inversion results of PLSR method and PCR method were better than those of SVR method.Comparing the inversion models of different types of soil and different soil parameters,the inversion results of pH value and total potassium content of paddy soil were stronger than those of purple soil,and the inversion effect of total potassium content was better than that of pH value.THese results of this study indicate that it is feasible to retrieve soil pH value and total potassium content by hyperspectral method.
作者
沈从旺
徐丽华
SHEN Cong-wang;XU Li-hua(College of Resources and Environment,Southwestern University,Chongqing 400715,China)
出处
《江苏农业学报》
CSCD
北大核心
2020年第1期92-98,共7页
Jiangsu Journal of Agricultural Sciences
基金
中央高校基本科研业务费专项(XDJK2016C083)
重庆市基础科学与前沿技术研究一般项目(cstc2016jcyjA0184)
国家自然科学基金项目(41671291)
关键词
支持向量回归
PH值
全钾含量
高光谱反演
support vector regression
pH value
total potassium content
hyperspectral inversion
