Author Affiliations
Department of Surveying and Mapping Engineering, School of Traffic & Transportation Engineering, Changsha University of Science and Technology, Changsha, Hunan 410004, Chinashow less
Fig. 1. Creating map area based on coordinate extreme values
Fig. 2. Established grid index diagram
Fig. 3. Distribution of seed points in small grid
Fig. 4. Quadric surface fitted by seed points in small grid
Fig. 5. Index grids
Fig. 6. Small grid in large grid
Fig. 7. Distances between other points and plane ΔACD
Fig. 8. Distances between other points and plane ΔADE
Fig. 9. Flow chart of improved filtering algorithm
Fig. 10. Distribution of feature points and ground points before filtering
Fig. 11. Ground point distribution after filtering
Fig. 12. Ground point distribution of elevation difference between real terrain and fitted terrain
Fig. 13. Distribution of feature points and ground points under threshold conditions
Fig. 14. Distribution of feature points and ground points of Sample41 before filtering
Fig. 15. Ground point distribution of Sample41 after filtering
Fig. 16. Distribution of elevation difference between real terrain and fitted terrain of Sample41
Fig. 17. Distribution of Sample41 feature points and ground points under threshold condition
Fig. 18. Ground point distribution after filtering by improved moving surface algorithm
Fig. 19. Distribution of ground points after filtering by classical moving surface algorithm
Test data | Sample | Topographical feature |
---|
Site1 | Sample11 | Vegetation and buildings on steep slopes | | Sample12 | Buildings and small objects on the ground | | Sample21 | Narrow bridge | Site2 | Sample22 | Bridge | | Sample23 | Complex buildings, discontinuous terrain | | Sample24 | Steep slopes and vegetation | Site3 | Sample31 | There is a low value noise point | Site4 | Sample41 | Discontinuous terrain | | Sample42 | High frequency terrain relief | | Sample51 | Vegetation on the slope | Site5 | Sample52 | Steep slope | | Sample53 | Discontinuous terrain | | Sample54 | Village | Site6 | Sample61 | Discontinuous steep slope | Site7 | Sample71 | Bridge | Site8 | No sample | Intermittent terrain, ridge |
|
Table 1. Test dataset and sample attributes released by ISPRS
Sample | Filtered data | Sample data point |
---|
| Ground point | Feature point |
---|
Ground point | a | b | e=a+b | Feature point | c | d | f=c+d | Filtered point | g=a+c | h=b+d | m=a+b+c+d |
|
Table 2. Definition of filter error
Testdata | Sample11 | Sample23 | Sample41 | Sample51 | Sample53 |
---|
Sample attribute | Vegetation and buildings on steep slopes | Complex building | Aggregate low value points | Low vegetation, steep slope, ridge | Intermittent terrain | Number of sample points | 38010 | 25095 | 11231 | 17845 | 34378 | Filtered feature point | 18576 | 13109 | 4908 | 12878 | 28324 | Filtered ground point | 19434 | 11986 | 6323 | 4967 | 6054 | Type I error /% | 17.80 | 16.37 | 13.43 | 9.73 | 14.50 | Type II error /% | 4.12 | 8.72 | 10.30 | 7.31 | 8.52 | Total error /% | 11.90 | 9.08 | 12.35 | 8.41 | 14.20 |
|
Table 3. Number of ground and non-ground points in the sample survey area and three types of error ratio
Data sample | Type of error | PTD algorithm | Morphological algorithm | Moving surface algorithm | Our algorithm |
---|
| Type I error | 15.96 | 21.97 | 21.52 | 17.80 | Sample11 | Type II error | 3.65 | 3.16 | 5.95 | 4.12 | | Total error | 10.76 | 17.36 | 14.87 | 11.90 | | Type I error | 12.08 | 13.30 | 18.39 | 16.37 | Sample23 | Type II error | 3.81 | 14.90 | 9.02 | 8.72 | | Total error | 8.22 | 14.10 | 14.72 | 9.08 | | Type I error | 8.58 | 12.53 | 12.23 | 14.50 | Sample53 | Type II error | 16.76 | 14.23 | 42.77 | 8.52 | | Total error | 8.91 | 12.60 | 17.71 | 14.20 | | Type I error | 7.15 | 22.43 | 1.84 | 1.87 | Sample61 | Type II error | 0.17 | 0.94 | 6.79 | 5.43 | | Total error | 6.91 | 21.68 | 2.01 | 1.99 |
|
Table 4. Comparison of accuracies of 4 filter algorithms%
Type of error | Improved moving surface algorithm | Classical moving surface algorithm |
---|
Type I error | 9.73 | 11.23 | Type II error | 7.31 | 9.34 | Total error | 8.41 | 10.27 |
|
Table 5. Statistics of three types of error of improved algorithm and classical algorithm for Sample51%