1. UMass Lowell Computer Science 91.503 Graduate Algorithms Prof. Karen Daniels Spring, 2005 Computational Geometry Overview from Cormen, et al. Chapter 33

2. Overview ä Computational Geometry Introduction ä Line Segment Intersection ä Convex Hull Algorithms ä Nearest Neighbors/Closest Points

3. Introduction What is Computational Geometry?

4. Computational Geometry Telecommunications Visualization Manufacturing ComputerGraphics Design Analyze Apply CAD

5. Sample Application Areas Computer Graphics Geographic Information Systems Robotics Bioinformatics Astrophysics MedicalImaging Telecommunications Data Mining & Visualization

6. Typical Problems ä bin packing ä Voronoi diagram ä simplifying polygons ä shape similarity ä convex hull ä maintaining line arrangements ä polygon partitioning ä nearest neighbor search ä kd-trees SOURCE: Steve Skiena’s Algorithm Design Manual (for problem descriptions, see graphics gallery at ) (for problem descriptions, see graphics gallery at http://www.cs.sunysb.edu/~algorith)

7. Common Computational Geometry Structures Voronoi Diagram Convex Hull New Point source: O’Rourke, Computational Geometry in C Delaunay Triangulation

8. Sample Tools of the Trade Algorithm Design Patterns/Techniques: binary searchdivide-and-conquerduality randomizationsweep-line derandomizationparallelism Algorithm Analysis Techniques: asymptotic analysis, amortized analysis Data Structures: winged-edge, quad-edge, range tree, kd-tree Theoretical Computer Science principles: NP-completeness, hardness Growth of Functions Summations Recurrences Sets Probability MATH Proofs Geometry Graph Theory Combinatorics Linear Algebra

9. Computational Geometry in Context TheoreticalComputerScience Applied Computer Science AppliedMath Geometry ComputationalGeometryEfficient Geometric Algorithms Design Analyze Apply

10. Sample of Supporting Algorithmic & Math Areas & Techniques ä Computational Geometry ä Convex hulls ä Visibility polygons ä Arrangements ä Mathematical Programming ä Linear programming ä Integer programming ä Lagrangian relaxation ä Upper, lower bounding ä Dynamic Data Structures ä Algorithm Design Patterns ä search space subdivision ä binary search ä divide-and-conquer ä sweep-line ä discrete-event simulation ä Algorithm Analysis Techniques ä Complexity Theory ä NP-completeness, hardness ä Discrete Math ä Minkowski sum ä Monotone matrices ä Lattices ä Set operations: union, intersection, difference

11. Line Segment Intersections (2D) Intersection of 2 Line Segments Intersection of > 2Line Segments

12. Cross-Product-Based Geometric Primitives source: 91.503 textbook Cormen et al. p0p0p0p0 p2p2p2p2 p1p1p1p1 (1) p1p1p1p1 p3p3p3p3 p2p2p2p2 (2) p2p2p2p2 p1p1p1p1 p3p3p3p3 p4p4p4p4 (3) Some fundamental geometric questions:

13. Cross-Product-Based Geometric Primitives: (1) source: 91.503 textbook Cormen et al. Advantage: less sensitive to accumulated round-off error p0p0p0p0 p2p2p2p2 p1p1p1p1 (1) 33.1

14. Cross-Product-Based Geometric Primitives: (2) source: 91.503 textbook Cormen et al. p0p0p0p0 p2p2p2p2 p1p1p1p1 (2) 33.2

15. Intersection of 2 Line Segments source: 91.503 textbook Cormen et al. p2p2p2p2 p1p1p1p1 p3p3p3p3 p4p4p4p4 (3) Step 1: Bounding Box Test Step 2: Does each segment straddle the line containing the other? 33.3 p3 and p4 on opposite sides of p1p2

16. Segment-Segment Intersection ä Finding the actual intersection point ä Approach: parametric vs. slope/intercept ä parametric generalizes to more complex intersections ä e.g. segment/triangle ä Parameterize each segment Intersection: values of s, t such that p(s) =q(t) : a+sA=c+tC a b c d L ab L cd a b c d L ab L cd A=b-a p(s)=a+sA q(t)=c+tC C=d-c source: O’Rourke, Computational Geometry in C 2 equations in unknowns s, t : 1 for x, 1 for y

17. Demo Segment/Segment Intersection http://cs.smith.edu/~orourke/books/CompGeom/CompGeom.html

18. Intersection of >2 Line Segments source: 91.503 textbook Cormen et al. Sweep-Line Algorithmic Paradigm: 33.4

19. Intersection of >2 Line Segments source: 91.503 textbook Cormen et al. Sweep-Line Algorithmic Paradigm:

20. Intersection of >2 Line Segments source: 91.503 textbook Cormen et al. Time to detect if any 2 segments intersect: O(n lg n) source: 91.503 textbook Cormen et al. 33.5 Note that it exits as soon as one intersection is detected. Balanced BST stores segments in order of intersection with sweep line. Associated operations take O(lgn) time.

21. Intersection of Segments ä Goal: “Output-size sensitive” line segment intersection algorithm that actually computes all intersection points ä Bentley-Ottmann plane sweep: O((n+k)log(n+k))= O((n+k)logn) time ä k = number of intersection points in output ä Intuition: sweep horizontal line downwards ä just before intersection, 2 segments are adjacent in sweep-line intersection structure ä check for intersection only adjacent segments ä insert intersection event into sweep-line structure ä event types: ä top endpoint of a segment ä bottom endpoint of a segment ä intersection between 2 segments ä swap order Improved to O(nlogn+k) [Chazelle/Edelsbrunner] source: O’Rourke, Computational Geometry in C

22. Convex Hull Algorithms Definitions Gift Wrapping Graham Scan QuickHullIncrementalDivide-and-Conquer Lower Bound in  (nlgn)

23. Convexity & Convex Hulls  A convex combination of points x 1,..., x k is a sum of the form  1 x 1 +...+  k x k where ä Convex hull of a set of points is the set of all convex combinations of points in the set. nonconvex polygon convex hull of a point set source: O’Rourke, Computational Geometry in C source: 91.503 textbook Cormen et al.

24. Naive Algorithms for Extreme Points Algorithm: INTERIOR POINTS for each i do for each j = i do for each j = i do for each k = j = i do for each k = j = i do for each L = k = j = i do for each L = k = j = i do if p L in triangle(p i, p j, p k ) then p L is nonextreme O(n 4 ) Algorithm: EXTREME EDGES for each i do for each j = i do for each j = i do for each k = j = i do for each k = j = i do if p k is not left or on (p i, p j ) if p k is not left or on (p i, p j ) then (p i, p j ) is not extreme O(n 3 ) source: O’Rourke, Computational Geometry in C

25. Algorithms: 2D Gift Wrapping ä Use one extreme edge as an anchor for finding the next  O(n 2 ) Algorithm: GIFT WRAPPING i 0 index of the lowest point i i 0 repeat for each j = i for each j = i Compute counterclockwise angle  from previous hull edge Compute counterclockwise angle  from previous hull edge k index of point with smallest  k index of point with smallest  Output (p i, p k ) as a hull edge Output (p i, p k ) as a hull edge i k i k until i = i 0 source: O’Rourke, Computational Geometry in C

26. Gift Wrapping source: 91.503 textbook Cormen et al. 33.9 Output Sensitivity: O(n 2 ) run-time is actually O(nh) where h is the number of vertices of the convex hull.

27. Algorithms: 3D Gift Wrapping CxHull Animations: http://www.cse.unsw.edu.au/~lambert/java/3d/hull.html O(n 2 ) time [output sensitive: O(nF) for F faces on hull]

28. Algorithms: 2D QuickHull ä Concentrate on points close to hull boundary ä Named for similarity to Quicksort a b O(n 2 ) Algorithm: QUICK HULL function QuickHull(a,b,S) if S = 0 return() if S = 0 return() else else c index of point with max distance from ab c index of point with max distance from ab A points strictly right of (a,c) A points strictly right of (a,c) B points strictly right of (c,b) B points strictly right of (c,b) return QuickHull(a,c,A) + (c) + QuickHull(c,b,B) return QuickHull(a,c,A) + (c) + QuickHull(c,b,B) source: O’Rourke, Computational Geometry in C finds one of upper or lower hull c A

29. Algorithms: 3D QuickHull CxHull Animations: http://www.cse.unsw.edu.au/~lambert/java/3d/hull.html

30. Algorithms: >= 2D Qhull: http://www.geom.umn.edu/software/qhull/ Convex Hull boundary is intersection of hyperplanes, so worst-case combinatorial size (not necessarily running time) complexity is in:

31. Graham’s Algorithm ä Points sorted angularly provide “star-shaped” starting point ä Prevent “dents” as you go via convexity testing O(nlgn) Algorithm: GRAHAM SCAN, Version B Find rightmost lowest point; label it p 0. Sort all other points angularly about p 0. In case of tie, delete point(s) closer to p 0. In case of tie, delete point(s) closer to p 0. Stack S (p 1, p 0 ) = (p t, p t-1 ); t indexes top i 2 while i < n do if p i is strictly left of p t-1 p t then Push(p i, S) and set i i +1 else Pop(S)  source: O’Rourke, Computational Geometry in C “multipop” p0p0p0p0

32. Graham Scan source: 91.503 textbook Cormen et al.

33. Graham Scan source: 91.503 textbook Cormen et al. 33.7

34. Graham Scan source: 91.503 textbook Cormen et al. 33.7

35. Graham Scan source: 91.503 textbook Cormen et al.

36. Graham Scan source: 91.503 textbook Cormen et al.

37. Algorithms: 2D Incremental ä Add points, one at a time ä update hull for each new point ä Key step becomes adding a single point to an existing hull. ä Find 2 tangents ä Results of 2 consecutive LEFT tests differ ä Idea can be extended to 3D. O(n 2 ) Algorithm: INCREMENTAL ALGORITHM Let H 2 ConvexHull{p 0, p 1, p 2 } for k 3 to n - 1 do H k ConvexHull{ H k-1 U p k } H k ConvexHull{ H k-1 U p k } can be improved to O(nlgn) source: O’Rourke, Computational Geometry in C

38. Algorithms: 3D Incremental CxHull Animations: http://www.cse.unsw.edu.au/~lambert/java/3d/hull.html O(n 2 ) time

39. Algorithms: 2D Divide-and-Conquer ä Divide-and-Conquer in a geometric setting ä O(n) merge step is the challenge ä Find upper and lower tangents ä Lower tangent: find rightmost pt of A & leftmost pt of B; then “walk it downwards” ä Idea can be extended to 3D. Algorithm: DIVIDE-and-CONQUER Sort points by x coordinate Divide points into 2 sets A and B: A contains left n/2 points B contains right n/2 points Compute ConvexHull(A) and ConvexHull(B) recursively Merge ConvexHull(A) and ConvexHull(B) O(nlgn) A B source: O’Rourke, Computational Geometry in C

40. Algorithms: 3D Divide and Conquer CxHull Animations: http://www.cse.unsw.edu.au/~lambert/java/3d/hull.html O(n log n) time !

41. Lower Bound of O(nlgn)  Worst-case time to find convex hull of n points in algebraic decision tree model is in  (nlgn) ä Proof uses sorting reduction: ä Given unsorted list of n numbers: (x 1,x 2,…, x n ) ä Form unsorted set of points: (x i, x i 2 ) for each x i ä Convex hull of points produces sorted list! ä Parabola: every point is on convex hull ä Reduction is O(n) (which is in o(nlgn))  Finding convex hull of n points is therefore at least as hard as sorting n points, so worst-case time is in  (nlgn) Parabola for sorting 2,1,3 source: O’Rourke, Computational Geometry in C

42. Nearest Neighbor/ Closest Pair of Points

43. Closest Pair source: 91.503 textbook Cormen et al. Goal: Given n (2D) points in a set Q, find the closest pair under the Euclidean metric in O(n lgn) time. Divide-and-Conquer Strategy: - -X = points sorted by increasing x - -Y = points sorted by increasing y - -Divide: partition with vertical line L into P L, P R - -Conquer: recursively find closest pair in P L, P R - -  L,  R are closest-pair distances - -  min(  L,  R ) - -Combine: closest-pair is either  or pair straddles partition line L - - Check for pair straddling partition line L - - both points must be within  of L - - create array Y’ = Y with only points in 2  strip - - for each point p in Y’ - - find (<= 7) points in Y’ within  of p

44. Closest Pair source: 91.503 textbook Cormen et al. Correctness 33.11

45. Closest Pair Running Time: Key Point: Presort points, then at each step form sorted subset of sorted array in linear time Like opposite of MERGE step in MERGESORT… source: 91.503 textbook Cormen et al. R L

46. Additional Computational Geometry Resources ä Computational Geometry in C ä second edition ä by Joseph O’Rourke ä Cambridge University Press ä 1998 Web Site: http://cs.smith.edu/~orourke/books/compgeom.html See also 91.504 Course Web Site: http://www.cs.uml.edu/~kdaniels/courses/ALG_504.html