1 / 43

Delaunay Triangulation

Delaunay Triangulation. Jyun-Ming Chen Reference: deBerg et al. Angle vector of T : sorted internal angles Angle-optimal triangulation T : A( T )  A( T ’ ) for all triangulations T ’ of P. Important Concepts. A( T ) < A( T ’). Important Concepts.

alena
Download Presentation

Delaunay Triangulation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Delaunay Triangulation Jyun-Ming Chen Reference: deBerg et al.

  2. Angle vector of T: sorted internal angles Angle-optimal triangulation T : A(T)  A(T’) for all triangulations T’ of P Important Concepts

  3. A(T) < A(T’) Important Concepts • Illegal edge, edge-flip and angle-vector • Legality test: [The correctness of this criterion follows from Thales’ theorem]

  4. p s q r Proof • We need to show that angle vector gets bigger if illegal edge is flipped • There are six angles in current configuration: • pqr + psr > 180 are the larger ones;the rest four are the smaller ones • Need to show that once flipped, all six new angles are larger than the four old smaller ones • The ordering of angle vector is determined by smallest angle

  5. qsp > qrp qsr > qpr qps > qpr Thales’ theorem q r p s

  6. pqs > prs sqr > rps qrs > prs q r p s

  7. Algorithm • Only flip if ijkl are convex (non-convex cannot flip) • It will terminate • Too slow to be interesting • How to come up with initial triangulation T ?!

  8. This was first suggested by Lawson(77) Incremental Algorithm

  9. Recursive Call Complication on presence of p-1, p-2,p-3

  10. p s q r Swap Test (Text, exercise 9.5) Note the triangle pqr is CCW!

  11. q1 q2 p s q r Swap Test (TRIPACK, Renka96) For numerical robustness (next page): Swap if sin12  - swtol Just check the sign!!

  12. [2] [3] Numerical Robustness 2 cocircular Almost cocircular 3 4 1 [do not flip – avoid cycling] [4] Don’t flip flip [1]

  13. Observation • Every new edge created due to insertion of pr is incident to pr This is also called the Bowyer-Watsonalgorithm

  14. Boywer-Watson Algorithm Bowyer-Watson triangulation: circum-circles that contain the new point, and the resulting triangulation

  15. 2 5 3 1 4 • Create DT(P) • Mark center of circum-circle carefully • Do the in-circle test by comparing distances • Perform all necessary LegalizeEdge steps -1 (i) (i) -2 (i) -3

  16. 2 5 3 1 4 -1 (0) (i) (iv) -2 -3

  17. 2 5 3 1 4 -1 (0) (iv) (iv) -2 -3

  18. 2 5 3 1 4 -1 (iii) (0) (i) -2 -3

  19. 2 5 3 1 4 -1 (ii) (0) -2 -3

  20. 2 5 3 1 4 -1 (0) (i) (iii) -2 -3

  21. 2 5 3 1 4 -1 (ii) (iii) -2 -3

  22. 2 5 3 1 4 -1 (0) (0) -2 -3

  23. 2 5 3 1 4 -1 -2 -3

  24. Edge swap complexity!?

  25. Choosing Initial Triangle • far away so that they don’t influence the DT of P • Don’t want to introduce huge coordinate (numerical errors)

  26. The Choice Suggested by Text • Choose them to contain P • Modify illegal edge test whenever p-i are involved • p-1 is outside any circles defined by P • p-2 is outside any circles defined by P{p-1} • p-3 is outside any circles defined by P{p-1,p-2}

  27. Case Analysis: Is pipj legal? [0] Polygon concave: legal [i] Both i&j negative: legal [ii] i,j,k,l positive: normal case [iii] Exactly one of (i,j,k,l) negative: choose the one with positive as legal [iv]Exactly two of (i,j,k,l) negatives: choose the one with smaller negative as legal [v] Exactly three of (i,j,k,l) negatives: cannot occur

  28. Logical Analysis [i,j,k,l] i<0  j<0 NOT( i<0  j<0) [i,j,k,l] two negative: ij must be one negative [i,j,k,l] all positive [i,j,k,l] one negative [i,j,k,l] three negative: impossible

  29. To locate the triangle containing pr Data structure D as described in text Point-Location Subproblem

  30. Alternate Point-Location Test Abstract of [Mucke et al. 96]

  31. Explanation from [Kallmann et al.03]

  32. Example p

  33. AP: EMST (exercise 9.11) • Q(n2) edges, standard MST algorithm O(n2) • EMST  DG • O(nlogn) algorithm to compute EMST for P • TSP …

More Related