Research On The Rapid Measurement Model Of Humic Acid Content From Weathered Coal Mine Based On Hyperspectral

As a large agricultural country,China’s agricultural production efficiency is directly related to the food problem of more than one billion people.In order to improve the efficiency of agricultural production,the market demand for fertilizers in my country is expanding,and the fertilizer manufacturing industry is developing rapidly.Mineral humic acid is often used as a raw material for fertilizers,which can not only improve fertilizer utilization,but also adjust the physical and chemical properties of the soil,making the soil more suitable for planting crops.At this stage,mineral-sourced humic acid is widely used,and its demand is high.However,the method of detecting the content of mineral-sourced humic acid before use is time-consuming,laborious and costly,and the use of chemical reagents will cause a certain degree of harm to the environment.In order to reduce the secondary pollution caused by chemical reagents and save time and effort to solve the problem of rapid element detection,many industries have begun to try to use spectroscopy to replace traditional artificial techniques.Research based on the combination of weathered coal hyperspectral and mineral source humic acid content,establish a hyperspectral characteristic band-a quantitative analysis index of mineral source humic acid content,which can build a rapid measurement model of mineral source humic acid content,realize rapid measurement of mineral source humic acid content,and optimize fertilizer quality control management to improve agricultural production efficiency,and also provide certain reference value and reference significance for the establishment of rapid measurement models of indicators in other fields.In this paper,coal samples from different origins are used as the research object,and the samples are ground to a particle size of 0.15 mm,and the humic acid content of its mineral sources is studied based on its hyperspectral characteristics.The total humic acid content of mineral-sourced humic acid in coal samples at home and abroad was tested by volumetric method,and SPSS 19.0 was used for mathematical statistical analysis.Secondly,the Field Spec 4 Hi-Res surface spectrometer was used to collect the hyperspectral reflectance of the mineral source humic acid samples in the dark room,and the RS~3and View Spec Pro software were used to save the original spectral data of the mineral source humic acid samples.With the help and support of software programs such as DPS,Matlab,Origin,etc.,the hyperspectral band is used as a variable parameter,and the humic acid content of the mineral source gives its influence to different bands according to the difference,established multiple linear regression(MLR),partial least square regression(PLSR),support vector regression(SVR)and BP neural network(BPNN)prediction models of mixed hyperspectral and mineral humic acid conten.Therefore,after the weathered coal is obtained,the hyperspectral technology can be used to quickly detect the content of humic acid contained in the mineral source.The conclusions of this study are as follows:1.Comparing the six pretreatment methods,the optimal pretreatment method for the humic acid spectrum of mineral sources is obtained.The hyperspectral reflectance data of coal samples acquired indoors with a ground feature spectrometer needs to be preprocessed first.For this,this article uses first-order derivative,second-order derivative,multivariate scattering correction(MSC),standard normal variable transformation(SNV),2nd-order-5-point SG least square smoothing filter(SG),and Y-gradient generalized least squares weighting algorithm(Y-GLSW)has a total of 6preprocessing methods.Support vector machine was used to test the ability of spectra to predict the humic acid content of mineral sources after pretreatment.After verification using support vector regression,it is proved by standard normal variable transformation,multivariate scattering correction,first-order derivative,2nd-order-5-point SG least square smoothing filter smoothing and Y-gradient generalized least squares weighting after preprocessing of the algorithm,the prediction effect of the established prediction model is relatively good.The coefficients of determination R~2are 0.821,0.815,0.771,0.758,and 0.654,which are 0.273,0.267,0.223,0.21 and 0.106 higher than the R~2of the original spectrum respectively;The spectroscopic data modeling based on the second derivative pretreatment has low predictive ability for the content of humic acid in the weathered coal,and the coefficient of determination R~2is 0.065.The standard normal variable transformation method is used to preprocess the weathered coal hyperspectral to predict the best humic acid content of the mineral source.2.Optimizing the best rapid measurement model for the content of humic acid from weathered coal mines among the 4 models.The continuous projection algorithm is used to screen out the spectral characteristic bands with a strong correlation between the hyperspectrum and the humic acid content of the mineral source,so as to improve the accuracy of later modeling.Based on the selected sensitive bands,a prediction model of multiple linear regression(MLR),partial least square regression(PLSR),support vector regression(SVR)and BP neural network(BPNN)of weathered coal hyperspectral characteristic band-mineral source humic acid content was established respectively.Studies have shown that multiple linear regression has a higher R~2than support vector regression,partial least square regression and BP neural network,which can reach 0.851,with higher stability and more accurate prediction.It is more suitable for predicting the humic acid content of weathered coal,and it is more effective to construct a rapid test model.Research has shown that the use of standard normal variable transformation spectral preprocessing method is used to preprocess the hyperspectrum of weathered coal,and the continuous projection algorithm is used to filter out the spectral characteristic bands with strong correlation between the hyperspectrum and the humic acid content of the mineral source.After that,using multiple linear regression modeling,the best content quick measurement model among the four models of the weathered coal hyperspectral characteristic band-mineral source humic acid content in this study was finally constructed,so as to realize the rapid measurement of humic acid content from weathered coal mine.3.It is clear that there are obvious spectral differences between weathered coal and lignite in the hyperspectral band.In the 350-600 nm range,the reflectance of the humic acid content of mineral sources is basically the same,and the spectral reflectance is low.The reflectivity of lignite in the600-2500 nm band is higher than that of weathered coal,but the trend is roughly the same.When faced with an unknown coal sample,it can be distinguished from weathered coal or lignite based on its hyperspectral characteristics.