近红外漫反射光谱检测土壤有机质和速效N的研究  

Near infrared diffuse reflectance spectra detection of soil organic matter and available N

作　　者：刘雪梅[1] 

机构地区：[1]华东交通大学土木建筑学院,南昌330013

出　　处：《中国农机化学报》2013年第2期202-206,共5页Journal of Chinese Agricultural Mechanization

基　　金：国家自然科学基金项目(41073060);江西省科技支撑项目江西省科技支撑项目(2009AE01603;2010EHB02000);铁路环境振动与噪声教育部工程研究中心资助

摘　　要：利用近红外漫反射光谱检测技术对土壤有机质和速效N含量进行了相关研究。通过自行设计的NIR光谱系统测定了150个土壤样品有机质和速效N。126个土壤样品用来建立校正集模型,其余24个用来验证模型的性能。采集完整土壤样品的近红外漫反射光谱,原始光谱经移动窗口平滑处理、SNV和一阶微分预处理后,分别采用最小二乘支持向量机(LS-SVM)和偏最小二乘法(PLS),建立土壤有机质和速效N含量的定量预测数学模型。结果表明采用一阶微分结合最小二乘支持向量机(LS-SVM)所建模型的预测效果较好,土壤有机质和速效N含量定量预测数学模型的决定系数分别为0.8255和0.8015,均方根误差分别为2.84和16.80。近红外漫反射光谱作为一种检测方法,可用于评价土壤有机质和速效N含量。Using near infrared diffuse reflectance nondestructive testing technology on soil organic matter and available N content were studied.Through the design of NIR spectroscopy system were determined in 90 samples of soil organic matter and available N.66 soil samples were used to establish the correct set of models,the remaining 24 is used to validate the model performance.Acquisition of intact soil samples by near infrared diffuse reflectance spectroscopy,the original spectra by moving window smoothing,vector normalization of SNV and first-order differ ential after pretreatment,respectively,by using least squares support vector machine（LS-SVM） and partial least squares（PLS）,establishment of soil organic matter and available N content quantitative prediction mathematical model.The results show that the first-order differential combined with partial least squares model better forecasting effect,soil organic matter and available N content quantitative mathematical model for prediction of correlation coefficients were 0.7977 and 0.8819,the root mean square error are 0.5967 and 2.1530.Near infrared diffuse reflectance spectroscopy as a detection method,can be used for the evaluation of soil organic matter and available N content.

关 键 词：近红外漫反射 土壤 有机质 速效N 

分 类 号：TH83[机械工程—仪器科学与技术] S15[机械工程—精密仪器及机械]