基于遥感影像的矿区植被恢复能力预测评估方法
By constructing a mining area-specific factor embedding layer and a Transformer architecture, the accuracy and generalization problems of the vegetation restoration model in mining areas during migration were solved, and adaptive parameter generation and rapid deployment were achieved, improving prediction accuracy and stability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA COAL INFORMATION TECH (BEIJING) CO LTD
- Filing Date
- 2026-02-06
- Publication Date
- 2026-07-17
AI Technical Summary
Existing vegetation restoration analysis models for mining areas based on remote sensing images suffer from decreased prediction accuracy and generalization ability when migrated to different mining areas. Furthermore, they require extensive data re-collection and parameter tuning, making it difficult to develop a general prediction tool that can be deployed on a large scale.
By collecting mining area-specific factors and constructing an embedding layer, mining area factor embedding vectors are generated. Combined with the encoder-decoder structure of the Transformer architecture, vegetation restoration features are extracted. A mining area factor branch network is established for nonlinear transformation to generate regulation coefficients. These coefficients are dynamically injected into the backbone model for multi-task learning to achieve adaptive parameter generation.
It enables adaptive adjustment of the model in different mining areas, reduces operation and maintenance costs, improves prediction accuracy and stability, is applicable to mining areas with multiple mineral types, supports rapid deployment and expansion, and enhances prediction robustness.
Smart Images
Figure CN121936734B_ABST