Fig. 1. Overall design of license plate recognition

Fig. 2. AOD-Net model

Fig. 3. K-value estimation model

Fig. 4. Comparison of AOD-Net model dehazing effect

Fig. 5. YOLOv5 network structure

Fig. 6. Mixed attention mechanism

Fig. 7. Channel attention module

Fig. 8. Spatial attention module

Fig. 9. Part of the network structure after integration into CBAM

Fig. 10. ASFF network structure

Fig. 11. Ghost convolution

Fig. 12. CCPD-Weather dataset

Fig. 13. Comparison of the images before and after haze enhancement

Fig. 14. Comparison of training model parameters

Fig. 15. Recognition results of license plate with clear image

Fig. 16. Test results of foggy images