dead simple buffer / difference approach

[smnorris]

Before improving snapping/road matching, perhaps a dead simple 7m buffer of higher priority roads and insertion of difference from lower priority roads will do the job. Provide a quick proof of concept.

[smnorris]

This works reasonably well to create a quick clean output that includes everything. If we wanted to extend this simple approach and further de-duplicate, I would consider:

• noding/cleaning the lower priority roads before inserting
• snapping intersections of lower priority roads to higher priority intersections at slightly higher tolerance before calculating difference/inserting
• after insert, compare lengths/counts of inserted features after being split up by the buffers and delete features that are obvious duplications (short relative to source feature, endpoints of features are still close (<10m? <20m? <50m?) to higher priority roads, high count of features inserted compared to the original feature)
• clean up results by snapping endpoints of retained features to closest higher priority feature (7.1m? 8m?)

[smnorris]

To get rid of obvious parallel lines, it should be straightforward to:

• find nearest neighbour of each lower priority line
• compare each lower priority line to its nearest neighbour, splitting them up into 10-50m chunks, comparing the distance from each lower priority vertex to higher priority vertex and seeing if the distance remains relatively constant. Breaking the higher priority lines at closest position to the endpoints of the lower priority lines and calculating the Hausdorf/Frechet distances may be better/same