Viewshed polygon optimisation

[tombh]

We currently store viewsheds using PolarSegments (the white example). When displaying in the browser we convert the PolarSegments to traditional polygons and union all the parts (the green example). Notice how it takes up 4 times as much storage. This is just one example, but it's likely that on average, PolarSegments are more efficient than polygons.

Using algorithms like Douglas-Peuker or Visvalingam we could simplify the polygon (the blue example), and even though in this case its byte size is equal to PolarSegents, it's likely that larger viewsheds enjoy reduced storage at the cost of some fidelity.

[ryan-berger]

I think that once we have 10x the number of segments, or even 100x the segments that we'll want something in-between the first and third. We can use the distance from the center to determine whether to simplify the polygon.

For example, within a radius of 1000 points from the point of view it may be the case that 10 segments or 100 segments make up .1 or .01 of the resolution of the underlying data. This would mean that each segment would have 10-100x more resolution than necessary. Simplifying a list of 10 or 100 segments that are exactly the same into a single block wouldn't lose resolution.

However, once you are outside of the radius where the arc sweep length matches the resolution of the underlying data you can't make any changes