Modified Model Of Grey Relational Estimation For Soil Organic Matter Based On Hyper-spectral

Soil organic matter(SOM)is one important component of soil.It is an important index to measure the soil fertility and monitor the crop growth.As for the traditional assay method for determination of soil organic matter,its accuracy is high,but the cycle is long and the cost is high,and it doesn't meet the need for rapid monitoring on a large scale.Hyper-spectral remote sensing technology has the characteristics of multiple bands,high resolution and large amount of data.It can monitor the change of soil organic matter content in real time and efficiently,so,it has some important significance for the development of fine agriculture.Zhangqiu District,Jinan City,Shandong Province,was selected as the research area.76 soil samples were collected.Based on their organic matter content and outdoor light abundance data,a modified model of grey correlation estimation for soil organic matter was established by using grey system theory based on hyper-spectral,and the preliminary estimation results of soil organic matter were revised.The main research contents and conclusions are as follows:(1)The characteristic index of soil organic matter in Zhangqiu district of Jinan city was determined.Five transformations of soil spectral reflectance,including first-order differential,square,reciprocal,logarithmic and square root were carried out.By comparing the soil spectral curves and soil correlation coefficient maps after different changes,it can be found that the fluctuation of soil spectral characteristic curves after first-order differential transformation was strong,and the correlation between spectral data and soil organic matter was obviously improved,and the effect of change was obviously better than the other four mathematical transformations.Therefore five bands of 544 nm,1469 nm,1656 nm,2059 nm and 2318 nm were selected as characteristic indicators according to the principle of maximum correlation from the soil spectral data after first-order differential transformation.(2)A modified model of grey correlation degree for soil organic matter estimation was established based on hyper-spectral.Based on the grey system theory,a modified model of hyper-spectral grey relational degree estimation of soil organic matter was established,and the preliminary estimation value of soil organic matter was modified.Firstly,the classical grey relational degree model was improved to obtain the weighted distance grey relational degree,and the weighted distance grey relational degree was used to estimate the soil organic matter content of the samples to be estimated preliminarily.Secondly,the preliminary estimates of soil organic matter were revised according to the modified model.Finally,for comparative analysis,the soil organic matter content was estimated by using support vector machine,multiple linear regression and BP neural network.The results showed that when estimating organic matter content based on grey model,the average relative error of those 16 samples was reduced from 17.396% to 3.801%,and the determinant coefficient was increased from 0.6413 to 0.9782.When estimating organic matter content by those three common models,the average relative errors of those 16 samples to be estimated were 8.373%,9.746% and 11.119%,and the determination coefficients were 0.8632,0.8433 and 0.7008.The study shows that the modified model of grey relational degree estimation of soil organic matter proposed in this paper can improve the estimation accuracy effectively and provide a new method for estimation of soil organic matter content based on hyper-spectral.