Method and system for remotely sensing farmland soil water based on crop physiological perception

CN122332833APending Publication Date: 2026-07-03WUHAN UNIV

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Applications(China)
Current Assignee / Owner: WUHAN UNIV
Filing Date: 2026-06-02
Publication Date: 2026-07-03

AI Technical Summary

Technical Problem

Existing remote sensing methods for farmland soil moisture retrieval suffer from severe vegetation interference, insufficient model generalization ability, and a lack of physical constraints, resulting in inadequate soil moisture retrieval accuracy and stability, making it difficult to meet the needs of high spatiotemporal resolution and large-scale monitoring.

Method used

A multi-source remote sensing inversion method based on crop physiological perception is adopted. The vegetation structure and physiological information are explicitly modeled through a multi-branch neural network structure. Combined with synthetic aperture radar and multispectral optical remote sensing data, a two-stage training strategy is adopted to train the auxiliary branch of vegetation structure and the main branch of soil moisture respectively. The physical interpretability and stability of the model are improved by using crop type, plant height and phenological stage information as constraints.

Benefits of technology

It significantly improves the accuracy and stability of soil moisture inversion, enabling high-resolution and high-precision soil moisture monitoring under dense crop canopy cover, and is suitable for fields such as precision agriculture, smart irrigation and agricultural ecological monitoring.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN122332833A_ABST

Patent Text Reader

Abstract

The application discloses a crop physiological perception-based multi-source remote sensing farmland soil water inversion method and system, and the method comprises the following steps: acquiring synthetic aperture radar data and multi-spectral optical remote sensing data and performing pretreatment to obtain target date remote sensing data and daily scale time series data; a multi-branch neural network model is constructed, including a soil water inversion main branch and three vegetation structure auxiliary branches corresponding to crop types, crop heights and phenological stage information respectively; a two-stage training strategy is adopted, the auxiliary branches are trained with measured vegetation parameters first, and then the main branch is trained with measured soil water content after the parameters of the auxiliary branches are frozen; pixel-level prediction is performed by using the trained model, and farmland soil water product is obtained in combination with land use data mask. By introducing crop physiological parameters as constraints, the application explicitly depicts the characteristics of vegetation structure, effectively reduces the interference of vegetation, and improves the precision, stability and generalization ability of farmland soil water inversion in crops with dense canopies.

Need to check novelty before this filing date? Find Prior Art