A small target detection method, system, device and medium
By dynamically adjusting the image detection kernel and using multi-level cross-channel attention extraction, the accuracy problem of small target detection at sea under complex sea conditions was solved, achieving high-precision detection in environments such as waves, changes in lighting, and sea fog.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HAIKOU SUB-BUREAU GUANGZHOU BUREAU EHV TRANSMISSION CO OF CHINA SOUTHERN POWER GRID CO
- Filing Date
- 2026-04-01
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies have low detection accuracy in detecting small targets at sea due to interference from complex sea conditions, and their robustness is poor, especially in environments with waves, changes in lighting, and sea fog.
By dynamically adjusting the threshold parameters and kernel size of the image detection kernel, combined with multi-channel stitching and multi-level cross-channel attention extraction, a trained small target detection model is used to detect small targets at sea, adaptively suppressing noise and extracting edge features.
It improves the robustness and accuracy of small target detection at sea, enabling accurate identification of small targets under complex sea conditions.
Smart Images

Figure CN122415976A_ABST