Fig. 1. Process framework

Fig. 2. Partial training data of motion blurred restoration. (a) Clear images; (b) partial enlarged images of Fig.2(a); (c) motion blurred images; (d) partial enlarged images of Fig.2(c)

Fig. 3. Structure of GAN

Fig. 4. Network structure of ViT-GAN. (a) Generator; (b) discriminator

Fig. 5. Network structure of MobileViT

Fig. 6. Network structure of MobileViT block

Fig. 7. Network structure of MV2 block

Fig. 8. Network structure of improved YOLOv8

Fig. 9. Structure of C2f before and after improvement. (a) Original C2f module; (b) C2f_MHSA module

Fig. 10. Structure of MHSA mechanism

Fig. 11. Improvement diagram of small object detection layer

Fig. 12. SIoU diagram

Fig. 13. Six types of small targets on transmission lines. (a) Insulator missing; (b) wire breakage; (c) damper; (d) suspension clamp; (e) strain clamp; (f) spacer

Fig. 14. Comparison of motion blurred restoration results of different algorithms

Fig. 15. Comparison of detection results. (a) Original YOLOv8; (b) improved YOLOv8