tag:blogger.com,1999:blog-2420860529344694449.post2938472616436836310..comments2024-03-25T02:37:02.982-07:00Comments on Lin.ear th.inking: Improvements to robustness in JTS Delaunay TriangulationDr JTShttp://www.blogger.com/profile/02383381220154739793noreply@blogger.comBlogger5125tag:blogger.com,1999:blog-2420860529344694449.post-74511767667600143252012-02-13T11:41:07.013-08:002012-02-13T11:41:07.013-08:00@lellela: Sorry, no idea about this.@lellela: Sorry, no idea about this.Dr JTShttps://www.blogger.com/profile/02383381220154739793noreply@blogger.comtag:blogger.com,1999:blog-2420860529344694449.post-4747299477626141272012-02-13T03:52:28.256-08:002012-02-13T03:52:28.256-08:00hi yeah, is it possible to use delaunay triangulat...hi yeah, is it possible to use delaunay triangulation maps on palm-print image for recogniion? can you please give solution for this question because i am doing my project on this .<br /><br />thankyou,<br />prashanthlellelahttps://www.blogger.com/profile/03841011569745736053noreply@blogger.comtag:blogger.com,1999:blog-2420860529344694449.post-4387432947691479942010-06-11T09:25:06.182-07:002010-06-11T09:25:06.182-07:00Actaully, make that 2x slower for using inCircle i...Actaully, make that 2x slower for using inCircle in DD versus double-precision.<br /><br />Still preferable to failing, but also a significant time penalty. The way to reduce this would be to introduce a fast filter to decide whether DD was necessary or not. I'm not sure what this would look like, though. If anyone has any ideas I'd be glad to hear them.Dr JTShttps://www.blogger.com/profile/02383381220154739793noreply@blogger.comtag:blogger.com,1999:blog-2420860529344694449.post-10810478916117023612010-06-11T08:18:46.125-07:002010-06-11T08:18:46.125-07:00Hi, Michael.
Yes, the dataset was posted on the J...Hi, Michael.<br /><br />Yes, the dataset was posted on the JTS list on April 11. Here's the link:<br /><br />http://sourceforge.net/mailarchive/forum.php?thread_name=1270982868.2207.23.camel%40braz-laptop&forum_name=jts-topo-suite-user<br /><br />I don't do anything special for 4 points aligned on a circle. I have tested this tho - the algorithm seems to be robust, but the results are non-determinate.<br /><br />I did some some performance testing of the DT using DD. It was only about 20% slower, I think - so well worth the cost.Dr JTShttps://www.blogger.com/profile/02383381220154739793noreply@blogger.comtag:blogger.com,1999:blog-2420860529344694449.post-42117033717942055722010-06-10T23:27:35.933-07:002010-06-10T23:27:35.933-07:00Hi,
Good work, Martin.
I had similar problems wit...Hi,<br /><br />Good work, Martin.<br />I had similar problems with my triangulation library, but I haven't get error for a long time (probably because I don't use it ;-)<br /><br />Do you know if the dataset which caused failure is available for tests ?<br /><br />Another point that worries me in my lib is that I do not detect 4 points strictly aligned or stricly lying on the same circle. A few month ago, I fixed some weird results I get while triangulating a regular grid, but I'm quite sure my code is not fully robust if alignment or co-circle points are found in the dataset. Do you use some tips to deal with these cases.<br /><br />You may remember some benchmarking I did a few month ago with double, DoubleDouble and some Arbitrary precision libraries. I get a coeff of 40x longer with DoubleDouble than with double (and 300x with BigDecimal). Do you confirm this results or do you get better ones ?<br /><br />Thanks again for your excellent work<br /><br />MichaƫlUnknownhttps://www.blogger.com/profile/18440967332483284330noreply@blogger.com