Dear Bin,
1) I read your paper and I found very rude from you using my algorithm
to aggregate lines without citing my work. Linear Strokes (look for
Robert Thomson) are a wider concept that searches generalisation (as
in Jiang B. and Claramunt C. (2004), A structural approach to model
generalisation of an urban street network). It can be applied not only
to streets, but also to river networks, for instance.
Robert's suggestion for replacing axial lines for linear strokes
(Thomson, 2003, bending the axial line) assumes a classic street
network based on the node-junction representation. The only know real
implementation of this approach is Porta et al (2006), which has a
clever and efficient algorithm that negotiates the best continuity.
Aggregating lines straight away from street segments (either axial
lines or road centre lines) using an "Angle of Deflection" was
proposed by me:
"The technical procedure to aggregate axial lines is based on the
"aggregation angle" or "angle of continuity", which is the angle
between the linear continuation of an axial line and the "real"
continuation provided by another axial line " (Figueiredo and Amorim,
2004, p5)
"The implementation of the aggregation process is based on the angle
between the lin-
ear continuation of an axial line and the \real" continuation provided
by another axial
line closer to one of its extremities. This angle of intersection is
called "angle of continu-
ity" " (Figueiredo and Amorim, 2005, p164)
"The software was designed to read any existent axial map (...), but
it can also be adapted to create continuity maps from GIS databases
(Figueiredo and Amorim, 2005, p166)"
Instead of using Porta's algorithm, which is better for road centre
lines, you apparently adopted my algorithm with a very wider angle (70
degrees). It happens that you may have created crazy curves when
better continuities were availalble. The relevance of the aggregation
procedure for your paper is crucial, as the "power laws" are created
by the generalisation process.
I would say that a proper citation would not reduce the merit of your
contribution, could would improved it. Note how one can cite properly
his colleagues (Pages 2 and 3):
http://eprints.ucl.ac.uk/archive/00002694/
2) Your power-laws are curves. There is few samples in your plots
where one could choose a cut-off point and find a tail that resembles
a straight line.
3) Instead of using your definition of n-neighbour clustering
coefficient, which is more appropriate for street networks, you used
the standard one. Street networks are grids, therefore we should look
a cycles with 3 steps (squares), not only 2 (triangles).
http://eprints.ucl.ac.uk/archive/00002694/
Best Regards,
Lucas
On 31/05/07, Bin Jiang <[log in to unmask]> wrote:
> Thanks all for the interesting discussion. Herewith my feedback:
>
> The log-log lines are not very straight, and in some case seem pretty
> curved. However the pattern (80/20 partition around the average m) we
> illustrated seems universal. With this pattern we do believe power-law
> there.
> > but this is curved throughout the length - very clearly so - I suspect a
> > simple power function which would still be interesting...
> >
> >
> >> One of the problems I had in the Physica A paper was that we do not know
> >> what a street is. There are several representations out there, but there
> >> is no uniquely accepted concept?
> >>
> >
> > Agreed - an argument for a (comparatively) well defined concept like the
> > axial line perhaps? :-)
> >
> Well, I cannot agree with the point. Through the experiments, we see the
> concept of streets based on perceptual grouping is pretty clearly
> defined, much better than axial lines. People may argue that the
> criteria of good continuity is a bit vague. In fact, we tried different
> threshold angles (actually a series from 20 to 90 degrees) for merging
> (or grouping) street segments to form individual streets, and found no
> big change in the illustrated pattern. One of my students Chengke Liu
> has testified three models: axial, stroke, named streets, and found the
> latter two are the best.
>
> Cheers.
>
> Bin
>
> --
> ------------------------------------------------
> Bin Jiang
> Department of Land Surveying and Geo-informatics
> The Hong Kong Polytechnic University
> Hung Hom, Kowloon, Hong Kong.
> Tel: (852) 2766 4335, Fax: (852) 2330 2994
> Email: [log in to unmask]
> ------------------------------------------------
>
--
Lucas Figueiredo
http://www.flickr.com/photos/lucasfigueiredo/
Mindwalk
http://www.mindwalk.com.br
|