一种基于自注意力机制的土壤水分降尺度方法
By employing a deep learning framework based on self-attention mechanism and multi-scale convolutional feature extraction, the shortcomings of remote sensing soil moisture downscaling methods in feature representation and spatial dependency modeling are addressed. This achieves high-precision spatial refinement of soil moisture images, improving prediction accuracy and information fusion efficiency in complex terrain areas.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JIANGXI PROVINCIAL LAND & SPACE SURVEY & PLANNING RES INST
- Filing Date
- 2025-09-04
- Publication Date
- 2026-07-17
AI Technical Summary
Existing remote sensing soil moisture downscaling methods are insufficient in terms of feature representation ability, spatial dependency modeling, and multi-source data fusion, making it difficult to meet the requirements for high-precision and high-resolution spatial distribution of soil moisture, especially in complex terrain areas where the prediction effect is poor.
We employ a deep learning framework based on self-attention mechanism, combining multi-scale convolutional feature extraction, cross-attention fusion, and Transformer encoder to construct an end-to-end soil moisture downscaling network, achieving high-precision spatial refinement of low-resolution soil moisture images.
It significantly improves the model's prediction accuracy and spatial resolution in complex terrain areas, enhances the fusion efficiency and prediction accuracy of multi-source information, and outputs high-resolution, structurally clear, and detail-rich soil moisture images.
Smart Images
Figure CN121074670B_ABST
Abstract
Citation Information
Patent Citations
CN112989286A
CN120411699A