Fig. 1. Flowchart of defect detection

Fig. 2. Preliminary segmentation steps and results of weld

Fig. 3. Angle model between point cloud normal vector and point surface

Fig. 4. Adaptive segmentation process

Fig. 5. Accurate segmentation of weld

Fig. 6. Improved region growing algorithm procedure

Fig. 7. Distance model

Fig. 8. Inspection platform for laser welding defect detection

Fig. 9. Battery defect data. (a)‒(c) Pitted defect battery, pitted welding defect joint, and pitted defect point cloud; (d)‒(f) raised defect battery, raised welding defect, and raised defect point cloud; (g)‒(i) no defective batteries, battery welds, and weld point clouds

Fig. 10. Segmentation result

Fig. 11. Segmentation results of different algorithms. (a) Images; (b) proposed method; (c) RANSAC; (d) DBSCAN

Fig. 12. Segmentation results of weld overlap defects. (a) RGB image; (b) elevation point cloud map; (c) defect segmentation result based on region growth algorithm; (d) defect segmentation result based on improved region growth algorithm

Fig. 13. Segmentation results of pit defects. (a) RGB image; (b) elevation point cloud map; (c) defect segmentation result based on region growth algorithm; (d) defect segmentation result based on improved region growth algorithm

Fig. 14. Method for measuring size of weld pit and weld bump defects. (a) Colorful images; (b) three-dimensional shape characteristics of defect area; (c) two-dimensional projections of defect area

Fig. 15. Measurement error