Hi,
Le 26/09/2012 19:44, Richard Weait a écrit :
I think that "drawing all of the nodes and points manually" is an
important difference, from a quality point of view. Each node or way
that you draw by hand, is carefully considered and placed, one at a
time. It isn't perfect; nothing is. I suggest that this leads to a
kind of automatic quality control, as the nodes and ways are placed.
You may draw carefully and at the wrong place at the same time. Drawing
manually is not a guaranty of right geometry. I (and many others in
France) used to draw buildings manually before june 2010 and the start
of the buildings-as-osm-files availability. This is time consuming and a
pain for both your eyes and wrist and fingers (and mouse too :-) ). So I
can't believe that after 30 minutes of such drawing your accuracy is
still at the top.
On the other side as explained by Christian, raw .osm files for
buildings are directly taken frow a vector source. When we display the
WMS version of the vector cadastre, we display exactly the same source.
As a consequence, it is quite easy to cross check .osm files once
displayed as a layer on the top of the WMS layer. As a matter of facts,
coordinates stored in the .osm file are the same as the ones used to
draw building polygons of the WMS layer : the source is the same, only
the way of displaying it differs (vector layer vs WMS layer). Drawing
manually can not produce such accurate geometry.
The goal of the vector import procedure is similar, use data from this
area, reconcile it carefully, include it in OpenStreetMap. The
intention is very good. But in execution, it is easier to miss a node
or way (or more than one) that needs to be refined before upload.
Again, it isn't perfect; nothing is. When you are considering
hundreds or thousands of nodes and ways at once, it becomes time
consuming to check them all.
Right. Done carefully, once again, integration of cadastre's buildings
*is* time consuming.
I hope that you'll find the above to be easy to agree with.
My conclusion, is this. The quality of the hand drawn nodes and ways
will be better because when we draw the nodes and ways by hand they
get more individual attention and care than when we start with a group
of nodes and ways from another file.
So that's why I think that it is different to trace by hand, vs.
vector import.
But what about the rules and edge cases?
I consider what you describe as the raster process to be an import
when the quantity of data is large. The raster process relies on an
external data source for a large quantity of information. If that
source is used without considering additional data sources, I think
that the classification as import is very clear.
On the other hand, I think that if the quantity of data is small, if
multiple sources are considered with appropriate weight, and if local
knowledge or an in-person survey is included as well? Then the
description might be closer to "really good mapping."
You ask what is the difference between the raster process and tracing
from aerial imagery alone. Good question. We haven't to this point
considered aerial tracing to be an import, but perhaps we should.
Perhaps the reason that tracing aerial imagery is not-an-import is
because it is transformative in a more obvious way? Tracing aerial
imagery transforms from a (rectified and positioned) picture of the
real world, to a vectorized and tagged abstraction. Tracing the
raster procedure, if I understand it correctly, transforms from a
raster version of one vectorized and tagged abstraction to second
vectorized and tagged abstraction.
Tracing the raster procedure transforms from a raster version of a
*paper map*, not taken from a vectorized source.
As mentioned earlier in this thread, French cadastre is still made of
rasterized old paper maps in 25% of our municipalities. Sometimes such
maps are geo-referenced by the Cadastre authorithy and we take it into
account. But in many cases each map is not georeferenced at all. The
contributor has to deal with it manually in JOSM. And you can have up to
dozens of maps for a single municipalities : a kind of puzzle.
See below [1] for a sample.
Before tracing from such maps you have to process a transformation that
is not a regular orthorectification (we do not use a DEM) but we more or
less try to translate + rotate + scale maps so that they fit with other
sources : bing imagery, osm data, survey points.
I can definitly not consider such workflow as part of an import. Even if
I can draw 1000 buildings in a single day on the top of cadastre maps.
As you can notice French cadastre is all but a single source in a single
format with a nation-wide repository. It is a compilation of formats
(vector & raster), projections (Lambert Conformal in best cases, no
projection at all in worst cases), procedures (from raw osm files to
manualy geo-referenced maps).
vincent
[1] : follow theses steps to display a sample raster cadastre map
1- go there : http://www.cadastre.gouv.fr/scpc/rechercherPlan.do#
2- in the 'Ville,Commune' input (on the left part of the page) type :
CERNAY-LES-REIMS
3- click on 'Rechercher' (white text on blue background)
4- click on 'Voir toutes les feuilles de la commune'
5- scroll down and click on 'Feuille 000 AB 01 - Commune :
CERNAY-LES-REIMS (51420)'
6- the map will open in a pop-up window. On the left panel you can play
with zoom & pan tools and see many buildings and street names.
_______________________________________________
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk
