Analysis Of Hyperspectral Prediction Model Of Ore Source Humic Acid

The mineral source humic acid can effectively promote the absorption of nitrogen,phosphorus,potassium and various trace elements of crops.It also has the effect of improving soil,improving soil microbial activity,improving crop resistance and improving fertilizer utilization rate of crops.The application prospect is very broad.At present,the mineral source humic acid has been industrialized,and the measurement of its humic acid content still remains in the traditional chemical measurement method,which is time-consuming and labor-intensive and costly.Rapid measurement of humic acid content in mineral source humic acid by using hyperspectral technology to achieve rapid testing has great significance for improving the determination speed of humic acid and promoting the development of humic acid industry.In this study,the research objects are selected from the mineral source humic acid samples and the samples were ground into four different particle sizes respectively of 0.15mm,0.18mm,0.25mm and 0.6mm.Fu.The spectrum of the four particle size mineral source humic acid samples was determined by Thermo Fisher's Nicolet^TMiN^TM10 fourier transform infrared spectrometer.The humic acid content of the mineral source humic acid sample was chemically determined by residue method in the laboratory.Based on the hyperspectral technique,the absorption spectrum characteristics and spectral information of the fertilizer are fully analyzed.And the spectral data is preprocessed by the Y-gradient generalized least squares weighting algorithm?Y-GLSW?.Use the correlation analysis method,continuous projection algorithm,random frog algorithm to screen out the sensitive bands.Use the support vector machine to establish predictive model of mineral source humic acid spectrum and humic acid content in mineral source humic acid under four particle sizes,and test and verify the established models by many multiple indicators.And then obtain the optimal band screening method and the optimal particle size of estimated model.The main conclusions are as follows:?1?Determined the spectral response of mineral source humic acid with particle size.The mineral source humic acid spectrum has strong volatility in the range of 680-1680 nm,and the second strong volatility is in the range of 2680-4000 nm.The spectral curve is stable in the range of 1680-2680 nm.The original spectrum curve of the mineral source humic acid and the spectral curve after pretreatment improved stability was gradually with the particle size decreased.The 0.15mm particle size mineral source humic acid has the highest spectral stability in this study.?2?Analysis of the effect of Y-GLSW pretreatment method on the mineral source humic acid.At the 0.6mm,0.25mm,0.18mm,0.15mm particle size level,the correlation coefficient values of the humic acid content and the spectrum pretreated by the Y-GLSW algorithm respectively are 0.4197,0.4897,0.4866,0.4747;The correlation coefficient values of the humic acid content and original spectrum respectively are 0.1979,0.3436,0.3572,0.3340.The correlation coefficient values of Y-GLSW pretreatment are higher than the original spectra and have better correction effects.?3?Optimize the optimal particle size of the mineral source humic acid model.Based on the correlation analysis method,successive projections algorithm method,random frog algorithm method,gray relational analysis method to establish the humic acid content estimation model,compare the four particle sizes of 0.6mm,0.25mm,0.18mm,0.15mm accuracy of the model is estimated,and the optimal particle size is 0.15 mm.The R²,RMSEC,RPD of the calibration set are:0.8695,1.3592,2.8301;0.9915,0.3387,10.4757;0.8036,1.6678,2.8549;0.4522,2.7851,1.8033.The R,RMSEP,RPD of the testing set are:0.6598,1.5888,2.8112;0.8269,1.4530,3.3070;0.6617,2.4773,2.9260;0.3532,2.7217,1.5979.In summary,the optimal particle size of this study is 0.15mm,and the optimal band screening method in the optimal particle size is successive projections algorithm method.?4?Optimal the optimal wavelength screening method for mineral source humic acid model.The wavelengths selected by the four wavelength screening methods are uniformly derived from the spectral fluctuation region,the spectral sub-fluctuation region and the spectral stability region,indicating that intervals all have influence on the wavelength selection.Compared with the humic acid models accuracy of 0.15 mm of the four wavelength screening methods,the continuous projection algorithm was the best method.The R²,RMSEC,RPD of the calibration set are:0.9915,0.3387,10.4757,and the R,RMSEP,RPD of the verification set are 0.8269,1.4530,and 3.3070.This model has capabilities of spectral analysis and spectral prediction.The model accuracy of the four wavelength screening methods is:continuous projection algorithm>random frog algorithm,correlation analysis method>gray correlation analysis method.In summary,the use of hyperspectral data to predict humic acid content of mineral source humic acids is feasible,providing a more intuitive technical method and theoretical support for the high-spectrum quality inspection of mineral source humic acid.