Author Affiliations
1School of Precision Instrument and Opto-Electronics Engineering, Tianjin University, Tianjin 300072, China2Key Laboratory of Opto-Electronic Information and Technology, Ministry of Education, Tianjin 300072, China3School of Information Engineering, Tianjin University of Commerce, Tianjin 300134, Chinashow less
Fig. 1. Flowchart of proposed algorithm
Fig. 2. Homologous point cloud data. (a) Bun000; (b) Bun045; (c) ArmadilloBack_0; (d) ArmadilloBack_30
Fig. 3. Cross-source point cloud data. (a) Bag_Kinect; (b) Bag_SFM; (c) Tsinghua gate_Lidar; (d) Tsinghua gate_SFM; (e) Life science building_Lidar; (f) Life science building_SFM
Fig. 4. Point clouds to be registered with different scaling factors. (a) Bunny, so=20; (b) Bunny, so=10; (c) Bunny, so=1.25; (d) Armadillo, so=20; (e) Armadillo, so=10; (f) Armadillo, so=1.25
Fig. 5. Registration results of various algorithms for Bunny point cloud under different scale factors. (a) EBABC-RS-IR; (b) ICP; (c) Scale-ICP; (d) CPD; (e) proposed algorithm
Fig. 6. Registration results of various algorithms for Armadillo point cloud under different scale factors. (a) EBABC-RS-IR; (b) ICP; (c) Scale-ICP; (d) CPD; (e) proposed algorithm
Fig. 7. Point cloud registration results under different noise. (a) Bunny, 20 dB; (b) Bunny, 25 dB; (c) Bunny, 30 dB; (d) Armadillo, 20 dB; (e) Armadillo, 25 dB; (f) Armadillo, 30 dB
Fig. 8. Relative initial state of cross-source point clouds to be registered. (a) Bag; (b) Tsinghua gate; (c) Life science building
Fig. 9. Local amplification effect of Bag registration
Fig. 10. Tsinghua gate point cloud registration results. (a) Main perspective; (b) prone perspective; (c) side perspective
Fig. 11. Life science building point cloud registration results. (a) Main perspective; (b) prone perspective; (c) side perspective
Point cloud | Image to be registered | Target /source point cloud | Number of points | Angle of view |
---|
Bunny | Bun000 | Target | 40256 | 45° | Bun045 | Source | 40097 | Armadillo | ArmadilloBack_0 | Target | 19283 | 30° | ArmadilloBack_30 | Source | 12150 | Bag | Bag_Kinect | Target | 11595 | Unknown | Bag_SFM | Source | 21495 | Tsinghua gate | Tsinghua gate_Lidar | Target | 33721 | Unknown | Tsinghua gate_SFM | Source | 971436 | Life science building | Life science building _Lidar | Target | 761729 | Unknown | Life science building _SFM | Source | 1813056 |
|
Table 1. Point cloud data information
Point cloud | so | Scale-ICP | CPD | Proposed algorithm |
---|
Bunny | 20 | 1.425×10-2 | 2.127×10-3 | 1.970×10-3 | 10 | 1.425×10-2 | 2.138×10-3 | 2.086×10-3 | 1.25 | 1.425×10-2 | 2.152×10-3 | 1.985×10-3 | Armadillo | 20 | 1.503×10-2 | 1.652×10-2 | 8.525×10-3 | 10 | 1.503×10-2 | 1.663×10-2 | 7.763×10-3 | 1.25 | 1.503×10-2 | 1.629×10-2 | 7.758×10-3 |
|
Table 2. RMSE for each registration algorithm with different scaling factors
Point cloud | so | Scale-ICP | CPD | Proposed algorithm |
---|
Bunny | 20 | 44.280 | 91.110 | 17.120 | 10 | 43.449 | 91.470 | 17.670 | 1.25 | 43.452 | 88.928 | 16.852 | Armadillo | 20 | 9.629 | 12.577 | 12.335 | 10 | 9.306 | 15.457 | 11.416 | 1.25 | 9.598 | 13.499 | 14.477 |
|
Table 3. Time for each registration algorithm with different scaling factors
Point cloud | so | 20 dB | 25 dB | 30 dB | No noise |
---|
Bunny | 20 | 8.229×10-3 | 4.594×10-3 | 3.799×10-3 | 1.970×10-3 | 10 | 7.949×10-3 | 7.176×10-3 | 3.223×10-3 | 2.086×10-3 | 1.25 | 8.358×10-3 | 4.704×10-3 | 3.073×10-3 | 1.985×10-3 | Armadillo | 20 | 1.010×10-2 | 8.815×10-3 | 8.034×10-3 | 8.525×10-3 | 10 | 1.032×10-2 | 1.129×10-2 | 8.155×10-3 | 7.763×10-3 | 1.25 | 1.184×10-2 | 8.804×10-3 | 8.312×10-3 | 7.758×10-3 |
|
Table 4. RMSE of the proposed algorithm under different noise
Point cloud | so | 20 dB | 25 dB | 30 dB | No noise |
---|
Bunny | 20 | 20.279 | 20.879 | 19.432 | 17.120 | 10 | 20.289 | 20.099 | 20.604 | 17.670 | 1.25 | 20.617 | 19.997 | 17.419 | 16.852 | Armadillo | 20 | 14.045 | 14.774 | 14.308 | 12.335 | 10 | 14.931 | 14.570 | 13.654 | 11.416 | 1.25 | 14.098 | 14.826 | 13.956 | 14.477 |
|
Table 5. Time of the proposed algorithm under different noise