Fig. 1. Welding seam tracking experimental device

Fig. 2. Weld surface condition and welding fixtures

Fig. 3. ROI image and grayscale distribution

Fig. 4. Weld identification process. (a) Mean filtering; (b) extract seed points; (c) screening weld area

Fig. 5. Distribution of characteristic parameters. (a) Gray value of weld area; (b) gray value of undetermined weld area; (c) gray value change in the welding direction of the weld area; (d) gray value change in welding direction of undetermined weld area; (e) gray gradient of weld area; (f) gray gradient of undetermined weld area; (g) width of weld area; (h) width of undetermined weld area

Fig. 6. Influence of parameters in the algorithm on recognition effect and running time. (a) Mean filter size; (b) minimum gray threshold; (c) maximum gray value difference; (d) minimum weld area width

Fig. 7. Weld recognition results of different gaps. (a) 0.05 mm; (b) 0.1 mm; (c) 0.2 mm; (d) 0.3 mm; (e) 0.4 mm

Fig. 8. Weld recognition results of different scanning speeds. (a) 100 mm/s; (b) 200 mm/s; (c) 300 mm/s; (d) 350 mm/s

Fig. 9. Common steel plate weld tracking results

Fig. 10. Results of weld tracking. (a) Original picture; (b) weld track identified by the algorithm and the enlarged image of local details

Fig. 11. Weld cross section

Fig. 12. Weld trajectory and deviation value