一种基于分层相关性回传的弱监督遥感旋转目标检测方法
By combining hierarchical correlation backpropagation and the rotation representative point loss function, the problem of rotating target detection in remote sensing images is solved, the detection speed and accuracy are improved, and end-to-end rotating target detection is realized.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIHANG UNIV
- Filing Date
- 2022-11-04
- Publication Date
- 2026-07-17
AI Technical Summary
Existing weakly supervised target detection techniques struggle to handle rotating targets in remote sensing images, and the testing time is long, making end-to-end processing difficult.
A weakly supervised remote sensing rotating target detection method based on hierarchical correlation backpropagation is adopted. The hierarchical correlation backpropagation of category is carried out through multi-instance learning branch to generate rotating pseudo-labels. The rotating detector is trained using the rotating representative point loss function and the rotating target detection results are output.
It effectively reduces edge blurring and category ambiguity of feature maps, significantly improves the overall processing speed of remote sensing image target detection, and realizes end-to-end rotating target detection.
Smart Images
Figure CN116071576B_ABST