An Improved Fruit Fly Optimization Algorithm For Least Squares Support Vector Machine Mod- El In Soil Moisture Retrieval

Soil moisture has been playing a very important role in the whole ecological environment, which is a vital parameter in hydrology, agriculture, meteorology and other fields, and can pro-vide information for global water cycle, hydrological modeling, etc. Therefore, the retrieval of soil moisture is quite a popular issue in recent years.First, we analyzes the microwave scattering characteristics of bare surface, and simulate the relationship among the backscattering coefficient, the incident wave, surface roughness, soil moisture and other parameters based on AIEM model. It is used as the sampling data of the later retrieval.Second, this paper describes the Statistical Learning Theory (STL), and uses the Least Squares Support Vector Machine (LSSVM) as the inversion algorithm. We find that the improp-er selection of the regularization parameter and kernel function parameter will cause a great de-cline of the retrieval accuracy.Consequently, the Fruit Fly Optimization Algorithm (FOA) is the first time used to opti-mize the regularization and kernel function parameters in LSSVM for soil moisture retrieval. And then an Improved Fruit Fly Optimization Algorithm (IFOA) is proposed to solve the local optimum problem by introducing a self-adaptive factor. The result shows that IFOA not only solves the local optimum problem, but also improves the retrieval accuracy. What’s more, we also analyze the robustness of this model.At last, the Michigan bare surface data is applied to verify the correctness of the algorithm in this paper.