Fig. 1. Illumination diagrams. (a) (b) Plane illumination diagrams; (c) (d) spherical illumination diagrams

Fig. 2. Schematic diagram of steel ball surface area division

Fig. 3. Highlight of same scratch defect by different lighting. (a) Effect of regional lighting; (b) effect of strip lighting

Fig. 4. Spatial omni-directional light source model

Fig. 5. Image processing. (a) Filtering and noise reduction; (b) image segmentation; (c) image difference (reverse color); (d) image fusion (reverse color); (e) defect extraction (reverse color); (f) defect classification

Fig. 6. Detection system and target detection steel ball. (a) Detection system; (b) target detection steel ball

Fig. 7. Annular lighting scheme and regional lighting scheme. (a) Example of annular lighting scheme; (b) example of regional lighting scheme

Fig. 8. Schematic diagram of illumination area of annular scanning illumination

Fig. 9. Schematic diagram of effective detection area overlay

Fig. 10. Influence of noise segmentation threshold T_d on defect extraction. (a) T_d=10; (b) T_d=50; (c) T_d=30

Fig. 11. Result of grid search optimal double threshold

Fig. 12. Detection results of micro defects on surface of steel balls. (a) (e) Shallow scratch defect; (b) (f) pit defect; (c) (g) striation defect; (d) (h) spot defect

Fig. 13. Comparison of defect detection effects in different positions of same defect. (a)-(c) Position 1；(d)-(f) position 2

Fig. 14. Comparison of defect highlighting ability between annular lighting and regional lighting. (a) Illumination experiment of same micro striation; (b) illumination experiment of same micro scratch

Fig. 15. Schematic diagram of illuminated area after lighting scheme optimization

Fig. 16. Comparison of same defect extraction effects before and after lighting scheme optimization. (a) Defect image information; (b) extraction effect before optimization; (c) extraction effect after optimization