Composite insulator slender target lightweight detection method based on LF-YOLO and thermal imaging
By adding C2PSA_LSKA attention and FreqFusion_BiFPN bidirectional multi-scale fusion to the YOLO backbone, combined with the Slim_Neck lightweight structure, the problems of contour breakage and boundary blurring in infrared composite insulator detection are solved, achieving efficient and stable target detection, which is suitable for real-time deployment in complex backgrounds.
CN122156586APending Publication Date: 2026-06-05ZHENGZHOU UNIVERSITY OF LIGHT INDUSTRY
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHENGZHOU UNIVERSITY OF LIGHT INDUSTRY
- Filing Date
- 2026-03-05
- Publication Date
- 2026-06-05
Smart Images
Figure CN122156586A_ABST
Abstract
The application relates to the technical field of infrared insulator detection, in particular to a composite insulator slender target lightweight detection method based on LF-YOLO and thermal imaging, which comprises the following steps: adding a C2PSA_LSKA attention fused with LSKA after a YOLO main stem SPPF, enhancing a global receptive field and highlighting weak edge features; using FreqFusion to reweight high-frequency and low-frequency information between each layer of the YOLO neck, and then performing bidirectional multi-scale fusion through BiFPN; in the YOLO neck, a lightweight GSConv is matched with VoVGSCSP to form a Slim_Neck, which significantly reduces the neck calculation and parameter scale while maintaining efficient multi-scale feature transmission. Under the premise of not significantly increasing the parameters and calculation, the application enlarges the effective receptive field, enhances the attention degree to key defect features, thereby reducing the missed detection and false detection, improving the precision rate, recall rate and overall reliability.
Need to check novelty before this filing date? Find Prior Art