Author Affiliations
1School of Air Transport, Shanghai University of Engineering and Technology, Shanghai 201620, China2School of Urban Rail Transit, Shanghai University of Engineering and Technology, Shanghai 201620, Chinashow less
Fig. 1. Flow chart of improved Poisson surface reconstruction algorithm
Fig. 2. Flow chart of normal estimation
Fig. 3. Normal vector fitting structure diagram
Fig. 4. Results of ambiguity of MC algorithm. (a) Two forms, (b) four results
Fig. 5. Structure diagram of different algorithms. (a) MC algorithm; (b) improved DC algorithm
Fig. 6. Comparison of filtering effect of the proposed algorithm. (a) Original point cloud dataset; (b) K=30; (c) K=50; (d) K=70
Fig. 7. Visualization of normal estimation. (a) (e)(i) Point cloud after preprocessing; (b)(f)(j) normal estimation of two traditional algorithms; (c)(g)(k) normal estimation of Ref. [18]; (d)(h)(l) normal estimation of improved algorithm
Fig. 8. Comparison of time complexity between four algorithms
Fig. 9. Comparison of surface reconstruction with four algorithms. (a)(e) Reconstruction of traditional Poisson algorithm; (b)(f) reconstruction of algorithm in Ref. [18]; (c)(g) reconstruction of greedy projection triangulation algorithm; (d)(h) reconstruction of improved algorithm in this paper
Fig. 10. Improved algorithm for surface reconstruction of different point cloud data. (a) Table model reconstruction; (b) pig model reconstruction; (c) horse model reconstruction
Method | Number of point clouds |
---|
Rabbit | Horse | Hand | Table | Pig |
---|
Before denoising | 35947 | 48485 | 327323 | 460400 | 502964 | After denoising | 31018 | 41977 | 285671 | 451410 | 439329 |
|
Table 1. Comparison of number of point clouds before and after denoising of different point clouds
Algorithm | Time /s |
---|
Rabbit | Horse | Hand | Table | Pig |
---|
Traditional Poisson | 26 | 35 | 153 | 217 | 146 | Greedy projectiontriangulation | 10 | 16 | 58 | 70 | 39 | Ref. [18] | 27 | 35 | 152 | 210 | 148 | Improved algorithm | 20 | 29 | 107 | 152 | 76 |
|
Table 2. Comparison of reconstruction time for different point cloud data
Algorithm | Number of model patches |
---|
Rabbit | Horse | Hand | Table | Pig |
---|
Traditional Poisson | 26985 | 36556 | 260869 | 382635 | 362543 | Greedy projectiontriangulation | 62427 | 82328 | 558510 | 746385 | 712476 | Ref. [18] | 27593 | 37386 | 270170 | 386528 | 375241 | Improved | 38712 | 46634 | 297823 | 396427 | 385894 |
|
Table 3. Patch number of the reconstructed model of four algorithms
Algorithm | Precision /mm | Completion /% |
---|
Traditional Poisson | 4.4 | 86.53 | Greedy projectiontriangulation | 4.2 | 91.48 | Ref. [18] | 4.4 | 88.72 | Improved | 4.1 | 92.81 |
|
Table 4. Reconstructed model accuracy of four algorithms