土壤速效磷可见-近红外光谱检测方法

Study on visible-near infrared spectroscopy for detection of available phosphorus in soil

土壤速效磷是影响农作物生长发育的重要指标,利用可见-近红外光谱技术对速效磷含量定量估测可为精准施肥提供重要依据。采集农田土壤样本可见-近红外光谱数据,土壤样本共179个。在原始光谱基础上采用Savitzky-Golay卷积平滑,一阶微分,二阶微分,标准正态变换,多元散射校正以及去趋势校正等单一及其组合对原始光谱数据进行预处理,然后将可见-近红外光谱分为2个波段范围(400~850 nm和950~1600 nm)并与全波段分别建立偏最小二乘回归和最小二乘支持向量机回归模型。结果表明Savitzky-Golay卷积平滑结合去趋势校正预处理效果最好,在此基础上,利用400~850 nm波段建立的最小二乘支持向量机模型取得了最佳效果,其模型验证集的决定系数为0.78,均方根误差为3.79mg/kg,相对分析误差为2.17。因此,采用最小二乘支持向量机回归建模法建立土壤速效磷的光谱定量分析模型,可实现土壤速效磷的定量估测。

The soil available phosphorus is an important index affecting the growth and development of crops,and the quantitative estimation of the available phosphorus content by using visible-near infrared spectroscopy can provide an important basis for accurate fertilization.The visible-near infrared spectroscopy data of farmland soil samples were collected,and there were 179 soil samples.On the basis of the original spectrum,the Savitzky-Golay convolution smoothing,first derivative,second derivative,standard normal transformation,multiple scattering correction and dislodge tendency were used for preprocessing.Then the visible-near infrared spectroscopy was divided into two wavelength range(400-850 nm and 950-1600 nm)and the partial least squares regression and least squares support vector machine regression models were established with the whole band.Results show that the pretreatment effect of Savitzky-Golay convolution smoothing combined with dislodge tendency was the best.On this basis,the least squares support vector machine model established in 400-850 nm achieved best results,the decision coefficient of validation set was 0.78,root mean square error was 3.79 mg/kg,relative analysis error was 2.17.Therefore,the spectral quantitative analysis model of soil available phosphorus was established by using the least-squares support vector machine regression model,and the quantitative estimation of soil available phosphorus was realized.