To
overcome the uncertainty in the spectral estimation of soil organic matter, the
hyper-spectral estimation model of soil organic matter content is established
using grey system theory. Firstly, after introducing the generalized greyness
of grey number, the properties of the generalized greyness are analyzed.
Secondly, the modeled samples are ranked in the smallest to the largest in
terms of soil organic matter content, the moving variance of the ranked
estimators is calculated, the greyness of the lower, value and upper domains of
the estimators is calculated based on the moving variance, and the new
estimators are constructed based on the greyness. The estimation model of soil
organic matter content is built and the estimation accuracy of the model is evaluated
using the mean relative error and the determination coefficient. Finally, the
model is applied to estimate soil organic matter content in Zhangqiu District
of Jinan, Shandong Province. The results show that the generalized greyness of
grey number can effectively represent the interval grey number, reduce the
random error and grey uncertainty of the estimation factor, and the accuracy of
the proposed estimation model and test accuracy are significantly improved,
where the determination coefficient R2 = 0.929 and the mean relative error MRE
= 6.830% for the 12 test samples. The results further enrich the grey system
theory, and provide a new way to modify the estimation factors and improve the
spectral estimation accuracy of soil organic matter.
