Estimation of soil alkali-hydrolyzed nitrogen content based on partial least squares regression
The estimation model of soil alkali-hydrolyzable nitrogen content based on indoor hyperspectral data was constructed,which provided a new method for rapid and accurate acquisition of nutrient information in soil.106 soil samples collected from Urumqi,Xinjiang were air-dried,ground and sifted,and the reflectance spectral data were collected indoors.The collected spectral data were pretreated by Savitzky-Golay filtering,first-order differentiation(FDR),continuum removal(CR)and multiple scattering correction(MSC).On this basis,continuous projection algorithm was used to screen the characteristic bands of the pre-treated data,and partial least squares regression was used to establish a hyperspectral analysis model for predicting soil alkali-hydrolyzed nitrogen content after different pretreatment.Coefficient of determination(R2),root mean square error(RMSE),relative analysis error(RPD)and mean relative error(MAE)were used as evaluation indexes of the model.The results showed that the prediction accuracy of the continuous removal treatment was the most prominent among the four pretreatment methods.R2,RMSE,RPD and MAE were 0.90,13.0,2.26 and 0.13,respectively.The linear regression equation of the model was y=0.9316x+8.763.Therefore,the continuous projection algorithm combined with partial least squares regression could be used to estimate the content of alkali-hydrolyzed nitrogen in soil in Urumqi.The results could provide a theoretical basis for rapid inversion of soil alkali-hydrolyzed nitrogen content with indoor hyperspectral data.
partial least squares regressionhyperspectral remote sensingspectral transformationestimation modelcontinuous projection algorithm
万方数据
维普
