Hyperspectral inversion for soil moisture and temperature based on Gaussian process regression

The soil moisture and temperature significantly influence the natural environment. Hyperspectral remote sensing can serve as a pivotal technique to monitor soil surface. However, modeling soil parameters encounter the following problems: hyperspectral data is high-dimensional and non-linear, and hyperspectral datasets are of limited size. In this paper, we derive a framework for inversion of soil moisture and temperature. First, wavelet transform is adopted that is able to extract the main structure of spectrum curve and reduce the dimensionality of the hyperspectral data. Then, Gaussian process regression (GPR), which is suitable for small sample data, is applied to predict the soil moisture and temperature. The experimental results show that our model outperforms other methods in estimating soil character.