Laser & Optoelectronics Progress, Volume. 60, Issue 24, 2415002(2023)
Wheel Tread Anomaly Detection Based on Attentional Reverse Knowledge Distillation
Fig. 1. Reverse knowledge distillation network structure based on attention
Fig. 2. Structure of encoder bottleneck
Fig. 3. Structure of decoder bottleneck
Fig. 4. Multi-scale feature fusion
Fig. 5. Attention mechanism
Fig. 6. SE module structure
Fig. 7. SGSE module structure
Fig. 8. Wheel tread image. (a)‒(c) Normal wheel tread image; (d) peeling defect image; (e) scratch defect image; (f) injury defect image
Fig. 9. Segmentation model training loss curve
Fig. 10. Prediction diagram of wheel tread image segmentation
Fig. 11. Detection view on RD model before wheel tread image segmentation
Fig. 12. Detection view on RD model after wheel tread image segmentation
Fig. 13. Comparison of detection view of wheel tread by reverse knowledge distillation method and the proposed method
Fig. 14. Comparison of detection view of wheel tread by different anomaly detection methods and the proposed method
|
|
|
|
|
Get Citation
Copy Citation Text
Rongrong Qin, xiaorong Gao, Lin Luo, Jinlong Li. Wheel Tread Anomaly Detection Based on Attentional Reverse Knowledge Distillation[J]. Laser & Optoelectronics Progress, 2023, 60(24): 2415002
Category: Machine Vision
Received: Mar. 7, 2023
Accepted: May. 6, 2023
Published Online: Dec. 4, 2023
The Author Email: Gao xiaorong (gxrr@vip.163.com)