Author Affiliations
1School of Architectural and Surveying and Mapping Engineering, Jiangxi University of Science and Technology, Ganzhou, Jiangxi 341000, China2College of Chinese and Asean Arts, Chengdu University, Chengdu, Sichuan 610106, Chinashow less
Fig. 1. Original point cloud
Fig. 2. Spatial distribution of noise. (a) Plane point cloud map; (b) distance map of neighbor points
Fig. 3. Flow chart of proposed algorithm
Fig. 4. Characteristic density distributions of 10 points in neighborhood
Fig. 5. Characteristic density distributions of 20 points in neighborhood
Fig. 6. Characteristic density distributions of 30 points in neighborhood
Fig. 7. Characteristic density distributions of 40 points in neighborhood
Fig. 8. Extracted results. (a) Original point cloud; (b) 10 neighbor points; (c) 20 neighbor points; (d) 30 neighbor points; (e) 40 neighbor points
Fig. 9. Experimental results and accuracy. (a) Trend of slope difference; (b) trend of accuracy difference; (c) number of ground point clouds; (d) error curves
Fig. 10. Extraction results of radius filter, voxel filter, and statistical filter
Fig. 11. Extraction results based on Method-Library. (a) After denoising in front view; (b) after denoising in side view
Fig. 12. Extraction results of large-area point cloud. (a) Before denoising in front view; (b) after denoising in front view; (c) before denoising in side view; (d) after denoising in side view
Point cloud | Number ofneighborpoints is 10 | Numberof neighborpoints is 20 | Numberof neighborpoints is 30 | Numberof neighborpoints is 40 | Original point cloud |
---|
MSE /m | 0.020843 | 0.019873 | 0.019687 | 0.019522 | 0.063345 | Range /m | 4.3340 | 4.7985 | 1.7899 | 2.6912 | 20.0210 | Total-error | 0.018476 | 0.012885 | 0.008622 | 0.008104 | - | Time /s | 16.235 | 18.397 | 21.565 | 27.382 | - | Point after denoising | 498379 | 494477 | 493651 | 492698 | 510519 |
|
Table 1. Extraction results under different numbers of neighbor points
Algorithm | Parameter | MSE /m | Total-error | Time /s | Number ofpoints afterdenoising |
---|
Radius filter | Radius(m) &number ofneighbor points | 0.4 & 4 | 0.069622 | 0.145738 | 88.637 | 435540 | 0.4 & 8 | 0.081447 | 0.253467 | 86.866 | 343263 | 0.8 & 4 | 0.058681 | 0.055329 | 81.427 | 499560 | 0.8 & 8 | 0.063917 | 0.071181 | 88.412 | 487163 | Voxel filter | Length ofvoxel /m | 0.5 | 0.184946 | 0.429362 | 68.346 | 117062 | 1.0 | 0.421697 | 0.470527 | 68.706 | 21615 | 1.5 | 0.731575 | 0.480870 | 69.402 | 42099 | 2.0 | 1.189826 | 0.484947 | 67.711 | 13226 | Statistical filter | Number ofneighbor points | 10 | 0.064307 | 0.071037 | 92.593 | 487328 | 20 | 0.064344 | 0.075085 | 116.735 | 484783 | 30 | 0.064735 | 0.077632 | 146.932 | 483097 | 40 | 0.063777 | 0.079753 | 188.853 | 481701 |
|
Table 2. Extraction results of other algorithms