基于连续统去除的土壤有机质近红外光谱敏感波段提取研究  

Study on the ertraction of sensitive band of soil organic matter near infrared spectrum based on continuum removal

作　　者：闫姗姗[1] 程旭[1] 宋海燕[1] 

机构地区：[1]山西农业大学工学院,山西太谷030801

出　　处：《山西农业大学学报（自然科学版）》2016年第1期72-76,共5页Journal of Shanxi Agricultural University(Natural Science Edition)

基　　金：国家自然科学基金项目(41201294);山西省科技攻关项目(20130313010-6)

摘　　要：土壤是一个复杂的三相集合体,由土壤中不同物质引起的谱带信息重叠现象非常严重,故通过适当的谱图预处理来提取其敏感波段显得尤为重要。本研究将连续统去除方法引入到土壤有机质敏感波段的提取中,分析了当土壤有机质含量变化时其谱图的变化规律,结果表明:采用连续统去除方法可以很好的提取土壤有机质敏感波段,且以提取的敏感波段600nm、900nm和2 210nm为中心,建立的土壤有机质模型可以较准确的预测土壤有机质含量,其所建模型中预测样本均方根误差MSE为0.286,相关系数R为0.979,均优于全波段所建模型中预测样本的均方根误差3.395和相关系数0.861。连续统去除方法可以很好的提取土壤有机质敏感波段,该研究对土壤有机质快速定量测试仪的研制具有重要意义。The soil is a complex three phase aggregate and the overlapping bands of information caused by its different substances is very serious.Therefore,it is important to extract the sensitive bands through appropriate spectrum processing.The continuum removal method was introduced to extract the sensitive wave bands of soil organic matter and analyse the changing rules of the spectra with the changing of soil organic matter content.In order to further verify that these bands are the sensitive bands of soil organic matter,we applied this method to the extraction of soil moisture sensitive bands,the results were close with previous proposed moisture sensitive wavelengths.Then the soil organic matter prediction model was established by using BP neural network,the 600 nm,900nm and 2210 nm as the center of sensitive wavelengths,the results were the root mean square error was 0.286 and the correlation coefficient was0.979,they were better than the whole band the root mean square error 3.395 and correlation coefficient 0.861.It indicated that it was feasible to extract the sensitive bands of organic matter by using the continuum removal method.It was of great significance to the development of rapid quantitative test instrument of soil organic matter.

关 键 词：土壤 有机质 连续统去除 敏感波段 神经网络 

分 类 号：S153.621[农业科学—土壤学]