A method for vehicle target recognition in a complex scene based on YOLO-FSG
By improving the YOLO-FSG model and utilizing the spatial frequency hybrid convolution module and the global-local adaptive module, the problem of low vehicle target recognition accuracy in complex scenarios was solved, achieving higher detection accuracy and robustness.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- DALIAN UNIV
- Filing Date
- 2026-04-13
- Publication Date
- 2026-07-17
AI Technical Summary
Existing vehicle target recognition methods have low recognition accuracy in complex scenes, especially under factors such as occlusion, changes in lighting, and background interference, which makes it difficult to effectively capture subtle features, resulting in a decrease in detection accuracy.
The YOLO-FSG model is adopted, and the C3K2 module in the backbone network is replaced with a spatial frequency hybrid convolutional module (SFHC), a P2 detection head is added, and a global local adaptive module (GLAM) is introduced to improve vehicle recognition accuracy.
It significantly improves the accuracy and noise resistance of vehicle target recognition, and can achieve accurate detection in complex lighting and multi-scale scenes. The evaluation indicators are improved by 3%-5%, surpassing other mainstream models.
Smart Images
Figure CN122415979A_ABST