Author Affiliations
1 Army Military Transportation University, Tianjin 300161, China2 Institute of Military Transportation, Tianjin 300161, Chinashow less
Fig. 1. Core idea of DBSCAN algorithm
Fig. 2. Dividing a three-dimensional voxel grid. (a) Three-dimensional voxel grid; (b) voxel cell
Fig. 3. Two-dimensional schematic of a layer of a voxel grid
Fig. 4. Simplified example of a merged cluster
Fig. 5. Original point-cloud
Fig. 6. Results using statistical outlier removal algorithm.(a) MeanK=10; (b) MeanK=30; (c) MeanK=50
Fig. 7. Results using radius outlier removal algorithm. (a) MinNeighbors=5; (b) MinNeighbors=10; (c) MinNeighbors=15
Fig. 8. Denoising results using VG-DBSCAN algorithm. (a) Eps=1, MinPts=10; (b) Eps=1, MinPts=15; (c) Eps=1, MinPts=20
Fig. 9. Local denoising results using VG-DBSCAN algorithm. (a) Before denoising; (b) after denoising
Fig. 10. Point-cloud-matching results after denoising. (a) Before matching; (b) after matching
Algorithm | Parameter | Point size | | Consuming time /ms |
---|
| | Original | After denoising |
---|
| MeanK=10 | | 38284 | 180.23 | Statistical outlier removal | MeanK=30 | 42618 | 37945 | 227.68 | | MeanK=50 | | 36579 | 288.85 | | MinNeighbors=5 | | 39744 | 465.37 | Radius outlier removal | MinNeighbors=10 | 42618 | 38543 | 466.91 | | MinNeighbors=15 | | 37349 | 471.15 | | Eps=1, MinPts=10 | | 38248 | 92.74 | VG-DBSCAN | Eps=1, MinPts=15 | 42618 | 37153 | 109.39 | | Eps=1, MinPts=20 | | 36142 | 124.69 |
|
Table 1. Denosing results using three algorithms
Denoising operation | Point size | Euclidean fitness score | Consuming time /s |
---|
| | Current frame | Previous frame |
---|
Without | 42618 | 45104 | 13.27145 | 66.752 | With | 36015 | 37549 | 6.12517 | 21.215 |
|
Table 2. Point-cloud-matching results before and after denoising