Fig. 1. YOLOv4 network structure

Fig. 2. PANet network structure. (a) FPN backbone; (b) bottom-up path augmentation

Fig. 3. CIOU schematic diagram

Fig. 4. Data augmentation

Fig. 5. Marking of insulator samples

Fig. 6. Mosaic data augmentation

Fig. 7. Convergence curve of loss function

Fig. 8. Simulation test results. (a) No background interference; (b)(c) with towers interference; (d) single target; (e) double target; (f) multi-target

Fig. 9. P-R curves of K-means++ YOLOv4 model. (a) Insulator identification; (b) insulator burst fault identification

Fig. 10. Comparison of performance indicators of different detection algorithms