Hello, I don't know if you have a clear theoretical idea of what you want to do.
If you work on roads, maybe you don't want to make them "smooth" but make them usable by realistic cars (ie no sharp turn, no turn with a radius minus than XX meters, so on...). For instance, you could use "coupoundCurve" (ie a mixture of line and curve) to represent your road, as in this paper (see page 9): The fitting process is page 10 and straightforward, you could use pure postgis - pl/pgsql. Wilkie D., Sewall J., Lin M. C., Lin M. C. « Transforming GIS Data into Functional Road Models for Large-Scale Traffic Simulation ». *IEEE Transactions on Visualization and Computer Graphics* [En ligne]. 2012. Disponible sur : < http://dx.doi.org/10.1109/TVCG.2011.116 > I'm guessing that depending on the target volume/processing time/result complexity , you can either solve the problem with "mathematical" processing or with an heuristic With "mathematical" processing you will consider your linestring as a multi-segment line, and fit to it a splin/polygon/clothoid/whatever. This would be easier with PL/R, as the fitting can become complex depending on the criteria (C1, radius, etc). with heuristic you will look for places where angles are too sharp and then correct it using local tricks/postgis curves. This would be easier with Pl/pgsql Cheers, Rémi-C
_______________________________________________ postgis-users mailing list [email protected] http://lists.osgeo.org/cgi-bin/mailman/listinfo/postgis-users
