If you can get a polygon of the parallel lines then you can try just the CenterLineReplacer. This should give a pretty good fit. There might be some issues at the ends though. But that would be my first approach.

@aron​ do you have an example dataset that you can share?

Hi @Mark Stoakes​ ,

This question originated from an idea I had where I was wondering what results I would get if I morphed together two datasets with different levels of detail. Take a coast line as example and imagine two official datasets, one in 1:150k and the other in 1:10k. Could these be combined to get a line in 1:60k that would look better and be more accurate than the 10k source run through a generaliser? The underlying assumption here is of course that both sources have been properly cartographed for their respective scales. What results would I get if I gave the computer the input from two separate sources where humans have applied their experience and knowledge, in stead of just one.

So far this is an interesting rabbit hole, I have not yet prepared any datasets to test this idea on.

(edit: I of course ment to generalize the 10k, not the 150k source)

I think that if we use @virtualcitymatt​'s idea of creating a polygon and extracting it's centerline you cold use HullAccumulator to create a Concave Hull, do a little cleanup and send the result to a CenterlineReplacer and then to Generalizer.

On the other hand, cartographic generalization shouldn't be done this way. You should generalize the 1:10k geometry to obtain a valid 1:60k coast line.