A bearing surface defect detection method based on a sparse perception multi-scale collaborative segmentation network
By constructing a sparse sensing multi-scale collaborative segmentation network (S3-Unet), the problems of missing small targets, interference from complex backgrounds, and overfitting with small samples in bearing surface defect detection are solved, achieving high-precision and high-efficiency defect detection.
Patent Information
- Application Number
- CN202610465447.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-10
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies for detecting defects on bearing surfaces suffer from problems such as missed detection of tiny targets, interference from complex backgrounds, and overfitting to small samples, making it difficult to achieve high-precision and high-efficiency defect detection.
We construct a sparse-aware multi-scale collaborative segmentation network (S3-Unet), which enhances multi-scale feature extraction capabilities, suppresses background noise, corrects feature misalignment, and improves detection accuracy and robustness by introducing a deformable sparse context aggregation module (DSCA), semantic gated skip connections (SGSC), and bidirectional feature refinement upsampling unit (BFRU).
On the self-built bearing end face and outer diameter dataset, the average intersection-to-union ratio (MIoU) reached 84.13% and 86.84%, respectively, and the average accuracy exceeded 90%, which significantly improved the detection accuracy and anti-interference ability, and was superior to the mainstream model.
Smart Images
Figure CN122415467A_ABST