Fig. 1. Overview of proposed method

Fig. 2. Structure diagram of YOLOv5s network

Fig. 3. Experimental platform for data collection

Fig. 4. Diagram of model training loss

Fig. 5. Results of indentation area edge detection

Fig. 6. Image preprocessing

Fig. 7. Four selected regions of ROI

Fig. 8. Edge processing

Fig. 9. Detection results of Hough circles

Fig. 10. Experimental results of test set

Fig. 11. Experimental results under different illumination treatments