Fig. 1. Schematic of region segmentation based on 3D shape matching

Fig. 2. Schematic of bidirectional linear octree data organization

Fig. 3. Comparison of the position of the boundary box and the original image

Fig. 4. Diagram of inconsistent point cloud parameters. (a) Point clouds data; (b)reorganized data; (c) curved surface

Fig. 5. Inconsistent estimated by angle of normal

Fig. 6. Effect of consistence segmentation

Fig. 7. Histogram of curvature of regular shape

Fig. 8. Effect of bidirectional linear octree data organization. (a) Bunny={BBoxi}; (b) Bunny: {Ncenterboxi}; (c) Dragon={BBoxi}; (d) Dragon: {Ncenterboxi}

Fig. 9. Comparison of density before and after data preprocessing. (a) Voxel; (b) box

Fig. 10. Boundary box for different objects in a multi-object scenario. (a) Boundary box for different objects; (b) original image

Fig. 11. Results of segmentation algorithms using Bunny data set. (a) Region growing; (b) RANSAC (plane); (c) RANSAC (sphere); (d) RANSAC (cylinder); (e) K-means; (f) SS

Fig. 12. Results of segmentation algorithms using Coal data set. (a) Region growing; (b) RANSAC (plane); (c) RANSAC (sphere); (d) RANSAC (cylinder); (e) K-means; (f) SS; (g) measured object

Fig. 13. Contrast of consistency among the regions

Fig. 14. Contrast of shape uncertainty among the regions

Fig. 15. Contrast of algorithm running time