一种基于语义细节解耦的作物杂草分割方法及系统
By adopting a crop and weed segmentation method based on semantic detail decoupling, the problems of poor small target recognition and unclear boundaries in UAV farmland scenarios are solved, achieving high-precision and stable crop and weed segmentation, which is adaptable to farmland scenarios of different datasets.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUNAN AGRI UNIV
- Filing Date
- 2026-04-28
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies for crop and weed segmentation in drone-based farmland scenarios suffer from poor small target recognition, unclear boundaries, and insufficient adaptability across datasets, resulting in inadequate segmentation accuracy and stability.
A crop and weed segmentation method based on semantic detail decoupling is adopted. A unified category space is formed through label mapping and preprocessing. Combined with multi-level feature extraction, top-down fusion, semantic-detail decoupling enhancement processing and multi-loss joint optimization mechanism, boundary supervision information is generated to improve the adaptability and accuracy of the segmentation model.
It improves the segmentation accuracy and boundary clarity of crops and weeds in drone-based farmland scenarios, enhances the model's adaptability to different farmland data and the consistency of segmentation results, simplifies the inference process, and facilitates practical deployment.
Smart Images
Figure CN122115875B_ABST