I'm trying to find the best way to expand a polygon to contain the most points. A simple greedy algorithm might be to expand with a 1km square, placing the square at the border in a location that captures the most points at that step. But what if I have multiple steps, it might make sense to place a couple empty squares so that the polygon reaches a hub of points. Is there a name for this algorithm? Or a way to implement it with shapefiles?
Asked
Active
Viewed 47 times
1
-
1Maybe a convex hull, but frankly I'm a little confused. Could you draw the polygon you're looking for? – David Eisenstat Mar 23 '21 at 01:39
-
I already have a polygon. Start from that polygon, what's the best path to enlargen it (say, by adding 1 by 1 squares contiguously to the border, so that I capture the most points. In the image above, it would likely be expanding straight down at the bottom in order to capture around 6 of the red points. I don't need to visit every point once in a path, I don't even need to hit every point. Just the best way to expand – Ben Hendel Mar 23 '21 at 05:08
1 Answers
0
Do you mean the convex hull (https://web.mit.edu/urban_or_book/www/book/chapter6/6.4.7.html) ?
Algorithms for solving the convex hull problem are in https://en.wikipedia.org/wiki/Convex_hull_algorithms
The convex hull is related to the travelling salesman problem (https://en.wikipedia.org/wiki/Travelling_salesman_problem) ?
https://en.wikipedia.org/wiki/Travelling_salesman_problem#Exact_algorithms and https://cs.stackexchange.com/questions/1749/dijsktras-algorithm-applied-to-travelling-salesman-problem/1805

ralf htp
- 9,149
- 4
- 22
- 34
-
It's not convex hull since I already have a polygon of coverage I'm starting from – Ben Hendel Mar 23 '21 at 05:09
-
Many convex hull algorithms start with an initial polygon, see i.e. Chan's algorithm https://en.wikipedia.org/wiki/Chan%27s_algorithm – ralf htp Mar 23 '21 at 16:34
-