How useful are current 3D city models?

Keynote given at the 102nd OGC Technical Committee meeting, at TU Delft, Delft, the Netherlands

Transcript

1. How useful are current 3D city models? Hugo Ledoux 102nd OGC Technical Committee TU Delft, the Netherlands 2017-03-20
2. Not only my work, but that of my research group 2
3. Not only my work, but that of my research group 3 •Mathematically easy but tools are needed From multiple representation to higher dimensional GIS x y z time scale
4. More and more 3D (open) data from cities 4 Whole North Rhine-Westphalia: 10M+ LoD2 buildings
5. More and more 3D (open) data from cities 5 New York City
6. More and more 3D (open) data from cities 6 Helsinki: with textures
7. More and more 3D (open) data from cities 7 Whole of the Netherlands: 10M+ LoD1 buildings + terrain, roads, forest
8. More and more 3D (open) data from cities 8
9. Why do cities reconstruct their 3D city model?
10. 10 solar potential shadow analysis noise modelling energy estimation bomb detonation wind turbulence pollutant tracking
11. 3D in the mainstream: NY Times shadow analysis 11 https://www.nytimes.com/interactive/2016/12/21/upshot/Mapping-the-Shadows-of-New-York-City.html
12. Environmental law: from 2018 in NL 12 Video “Stip op de horizon: Dynamisch 3D model” from Municipality of The Hague
13. How are geometric errors affecting applications?
14. Consequences of geometric errors in 3D city models 14 1. for “simple” visualisation 2. for solar potential 3. for volume calculation 4. for advanced simulations I ignore the semantic here
15. Visualisation duplicated surfaces == annoying 15
16. Visualisation duplicated surfaces == distracting 16
17. Visualisation duplicated surfaces == distracting 17
18. Visualisation wrong orientation of surfaces (red ones) 18
19. Visualisation wrong orientation == missing faces 19
20. Solar potential wrong orientation == no potential assigned 20
21. Volume calculation tiny problems == impossible to calculate 21
22. Volume calculation tiny problems == impossible to calculate 22
23. Volume calculation big problems == also impossible to calculate… 23
24. Volume calculation superstructures not topologically connected 24
25. Volume calculation superstructures not topologically connected 25
26. Volume calculation superstructures not topologically connected 26
27. Advanced simulations wind comfort for pedestrians 27 Figure from CADFEM.de
28. Advanced simulations 3D CFD methods: tetrahedralisation of volume 28 computational fluid dynamics Figure from CADFEM.de
29. Advanced simulations very strict input requirements 29 Figure from CADFEM.de • no holes • no intersections • triangles must have certain shape (no slivers) • a “perfect” surface is thus required • in practice, that still means several hours of (semi-)manual repair
30. Do current 3D city models often contain these geometric errors?
31. tl;dl: YES.
32. Our methodology 37 CityGML datasets from 9 countries 40 million polygons in 3.6 million buildings (Solids + MultiSurfaces) our software: val3dity 32
33. Error codes 33 Solid + MultiSolid CompositeSurface MultiSurface LinearRing 101 TOO_FEW_POINTS 102 CONSECUTIVE_POINTS_SAME 103 NOT_CLOSED 104 SELF_INTERSECTION 105 COLLAPSED_TO_LINE Polygon 201 INTERSECTION_RINGS 202 DUPLICATED_RINGS 203 NON_PLANAR_POLYGON_DISTANCE_PLANE 204 NON_PLANAR_POLYGON_NORMALS_DEVIATION 205 INTERIOR_DISCONNECTED 206 HOLE_OUTSIDE 207 INNER_RINGS_NESTED 208 ORIENTATION_RINGS_SAME Shell 301 TOO_FEW_POLYGONS 302 NOT_CLOSED 303 NON_MANIFOLD_VERTEX 304 NON_MANIFOLD_EDGE 305 SEPARATE_PARTS 306 SELF_INTERSECTION 307 POLYGON_WRONG_ORIENTATION 308 ALL_POLYGONS_WRONG_ORIENTATION Solid (with 2+ Shells) 401 SHELLS_FACE_ADJACENT 402 SHELL_INTERIOR_INTERSECT 403 INNER_SHELL_OUTSIDE_OUTER 404 INTERIOR_OF_SHELL_NOT_CONNECTED For CompositeSurface these 3 are excluded LinearRing Polygon Point Solid ultiSurface CompositeSurface Shell CompositeSolid LinearRing Polygon Point Solid MultiSurface CompositeSurface Shell CompositeSolid
34. Summary of results 34 • No CityGML dataset is 100% valid, not even LoD1 • Many simple errors, eg: • repeated vertices • non-planar polygons (most common error!) • Notice that these are often not visible
35. Summary of results 35 • No CityGML dataset is 100% valid, not even LoD1 • Many simple errors, eg: • repeated vertices • non-planar polygons (most common error!) • Notice that these are often not visible
36. Level of detail Primitive ID(a) Geometric validation Semantics(f) Schema 1xx 2xx 3xx 4xx Total(b) LOD1 Solid 1 0 0 0 0 0 – 7 2 0 0 0 0 0 – 3 3 0 0 0 0 0 – 3 4 0 0 0 0 0 – 3 5 0 0 0 0 0 – 7 6 0 0 0 0 0 – 7 7 0 0 0 0 0 – 3 8 0 0 0 0 0 – 3 9 0 0 0 0 0 – 7 LOD2 MultiSurface 10 1 4 – – 5 (e) 3 11 0 0 – – 0 0 3 12 2 21 – – 23 45 3 13 10 2 – – 12 4 7 14 0 1 – – 1 12 3 15 0 9 – – 9 2 3 16 4 8 – – 12 1 7 17 5 0 – – 5 5 7 18 0 0 – – 0 4 7 19 0 0 – – 0 1 7 20 0 4 – – 4 6 7 21 0 1 – – 1 3 7 LOD2 Solid 22 0 42 58 0 100 – 3 23(c) – – – – – – 7 24 0 31 1 3 35 – 7 25 4 0 16 2 22 – 3 26(c) – – – – – – 7 27 22 17 50 0 89 – 3 LOD2 MultiSurface and Solid(d) 28 0 42 1 1 44 0 3 29 2 35 54 0 92 4 3 30 0 10 0 1 11 2 3 31 0 0 0 0 0 0 3 LOD3 MultiSurface 32 2 13 – – 15 54 3 33 6 5 – – 11 23 3 34 8 10 – – 19 45 3 35 5 0 – – 5 34 3 36 0 0 – – 0 1 7 LOD4 Solid 37 0 0 3 0 3 68 3 (a) % of invalid actually all between 99.5% and 99.9%
37. Level of detail Primitive ID(a) Geometric validation Semantics(f) Schema 1xx 2xx 3xx 4xx Total(b) LOD1 Solid 1 0 0 0 0 0 – 7 2 0 0 0 0 0 – 3 3 0 0 0 0 0 – 3 4 0 0 0 0 0 – 3 5 0 0 0 0 0 – 7 6 0 0 0 0 0 – 7 7 0 0 0 0 0 – 3 8 0 0 0 0 0 – 3 9 0 0 0 0 0 – 7 LOD2 MultiSurface 10 1 4 – – 5 (e) 3 11 0 0 – – 0 0 3 12 2 21 – – 23 45 3 13 10 2 – – 12 4 7 14 0 1 – – 1 12 3 15 0 9 – – 9 2 3 16 4 8 – – 12 1 7 17 5 0 – – 5 5 7 18 0 0 – – 0 4 7 19 0 0 – – 0 1 7 20 0 4 – – 4 6 7 21 0 1 – – 1 3 7 LOD2 Solid 22 0 42 58 0 100 – 3 23(c) – – – – – – 7 24 0 31 1 3 35 – 7 25 4 0 16 2 22 – 3 26(c) – – – – – – 7 27 22 17 50 0 89 – 3 LOD2 MultiSurface and Solid(d) 28 0 42 1 1 44 0 3 29 2 35 54 0 92 4 3 30 0 10 0 1 11 2 3 31 0 0 0 0 0 0 3 LOD3 MultiSurface 32 2 13 – – 15 54 3 33 6 5 – – 11 23 3 34 8 10 – – 19 45 3 35 5 0 – – 5 34 3 36 0 0 – – 0 1 7 LOD4 Solid 37 0 0 3 0 3 68 3 (a) % of invalid rather good, but it’s “easy”
38. Level of detail Primitive ID(a) Geometric validation Semantics(f) Schema 1xx 2xx 3xx 4xx Total(b) LOD1 Solid 1 0 0 0 0 0 – 7 2 0 0 0 0 0 – 3 3 0 0 0 0 0 – 3 4 0 0 0 0 0 – 3 5 0 0 0 0 0 – 7 6 0 0 0 0 0 – 7 7 0 0 0 0 0 – 3 8 0 0 0 0 0 – 3 9 0 0 0 0 0 – 7 LOD2 MultiSurface 10 1 4 – – 5 (e) 3 11 0 0 – – 0 0 3 12 2 21 – – 23 45 3 13 10 2 – – 12 4 7 14 0 1 – – 1 12 3 15 0 9 – – 9 2 3 16 4 8 – – 12 1 7 17 5 0 – – 5 5 7 18 0 0 – – 0 4 7 19 0 0 – – 0 1 7 20 0 4 – – 4 6 7 21 0 1 – – 1 3 7 LOD2 Solid 22 0 42 58 0 100 – 3 23(c) – – – – – – 7 24 0 31 1 3 35 – 7 25 4 0 16 2 22 – 3 26(c) – – – – – – 7 27 22 17 50 0 89 – 3 LOD2 MultiSurface and Solid(d) 28 0 42 1 1 44 0 3 29 2 35 54 0 92 4 3 30 0 10 0 1 11 2 3 31 0 0 0 0 0 0 3 LOD3 MultiSurface 32 2 13 – – 15 54 3 33 6 5 – – 11 23 3 34 8 10 – – 19 45 3 35 5 0 – – 5 34 3 36 0 0 – – 0 1 7 LOD4 Solid 37 0 0 3 0 3 68 3 (a) % of invalid some datasets couldn’t be read high % of invalid
39. Most common error: non-planar surface 39
40. How to improve the situation?
41. 1. Improve awareness most practitioners are not aware of the rules 41
42. 1. Improve awareness most practitioners are not aware of the rules 42
43. 1. Improve awareness most practitioners are not aware of the rules 43 software vendors are not either
44. 2. Include validation support in software or use the open-source software 44
45. 3. Modify software with constraints TU Delft’s 3dfier is only but one example 45
46. 4. Work on automatic repair funding a research team is a great way to start! 46 Surface reconstruction:
47. thank you. Hugo Ledoux h.ledoux@tudelft.nl 3d.bk.tudelft.nl/hledoux 3d.bk.tudelft.nl/code/ Open-source software