I can try and write some script that creates these tiles.
This script would be reusable if/when we get an updated version of this file.
I know this file change once in a while, when municipalities are formed or merged one into another.
If jpg/jgw combination is the standard, then we could use this script in future imports as well.
My biggest problem right now is: what is the naming scheme for tiles which Potlatch understands?
I managed to produce beautiful tiles which can be overlayed on Google Maps, but not in Potlatch.
Click on the arrow next to “Background”
Click on the “Edit…” button. You’ll get a “Background imagery” popup window.
Click on “Add” (bottom right), a new line will be added, and edit it as follows:
I have played a little with automatic vectorising before and I decided to try my old vectoriser on the .jpg file with the borders. The results are very bad, probably useless, but if you’re curious, I have put up some files to look at:
These results are so bad because the parameters are adjusted for a different task (I had building outlines at very very high resolution, so the vectoriser ignores short segments, and tries to find right angles everywhere, even if there aren’t any). I will play with the parameters a little tomorrow, today I spent all the time trying to get the EPSG:2039 projection right.
Dimka, could you ask MOIN for an image file without the names?
Sometimes the names are in the way, like in Abu Gosh / Kiryat Yearim.
We could trace on the nameless file, and use the names from the named file, switching between the2 overlayes.
The nameless file could be used by balrog-kun to get better results.
I now basically removed all of the building-related postprocessing and now the results are a little better, I don’t think it’s possible to do this better except for manual tracing.
Note that the blue lines have a width of 7 pixels, so they are polygons and it’s difficult to calculate only the centrelines (I think it’s impossible).
When the administrative borders run along streams (e.g. Beit Gamliel). What is more accurate, the borders based on the raster image you got, or the natural borders. And if the latter is true, should we correct them?
AFAIK, officially, the raster image is more accurate.
When streams change course, it takes time for the MOIN to do new surveys and update the administrative borders. If at all…
The raster image is clearly off by a few meters in some obvious places (like border between two cities which should pass in the middle of a highway). The lines on the raster are too thick and in many cases an adjustment is needed based on common sense (for example when a city border is on top of a residential area clearly belonging to the city, the border should be adjusted to be just outside the area).
I don’t know if the administrative borders are supposed to coincide with natural boundaries or just be parallel to them. I think that in case of streams it doesn’t really matter.