Very easy question for you (close a sort of perimeter)

Hi guys

I have a sort of "perimeter" (this perimeter could be different, with more openings instead of 3 of the image). I would close these "openings", but without the Extend command (these openings could have different length). Is there someone that could help me?

Attachments:

Screen.png, 30 KB

Replies to This Discussion

Permalink Reply by Landscape on October 11, 2012 at 4:27am

This is my new approach, I put a "point on a curve" ... is there the same problem????

Attachments:

Screen.png, 22 KB

Permalink Reply by Raul Matagringo on October 11, 2012 at 5:25am

the problen is that you cant order the points.

with Andrews solutionn wont work because L(side) > L(gap)

with my solution wont work because some the lines ftom "G" to the outer points will intersect

i dont care about L(side) > L(gap) and direction of curves but when the lines intersect i cant order the ventrices

Permalink Reply by Landscape on October 11, 2012 at 5:39am

mmmm ... and isn't possibile to "rename" pairs of points in clockwise? (maybe is what u did) ...

Permalink Reply by Raul Matagringo on October 11, 2012 at 8:16am

in my solution i reorder the points clockwise.

My test:

relevat points with index and the kind of center

my lines from center to the points

IMPORTANT: as you can see NO COLLISION between the lines. Now i can duild the angle (i check angle with a vec cross prod). I sort the angles and the node index

now i can pass the point as ventrices to a polyline component.

and at the end the polyline in green

Example of a bad figure

Why bad?

Because i can order the points in the marked area. The angle method is not working enymore.

Point 8 and 9 are ok. Point 6 comes now before the 9.

Point 14 is ok. But point 15 comes before. And so on...

I have no idea how to solve this

Permalink Reply by Landscape on October 11, 2012 at 8:31am

OK, now I understand. A solution could be generate 2 kind of center points instead one. But it's hard!!

Permalink Reply by Raul Matagringo on October 11, 2012 at 8:50am

no, there arent 2 center points. This kind of notch builded by 8/9/14/6 is the problem. Simplified: when the line from center intersect a edge line. You can see in the last picture. Points 9,14,6,15 are destroying the order.

Permalink Reply by Landscape on October 11, 2012 at 9:04am

I understand there is one central point, but I suppose that if the complex area will be subdivided in 2 o 3 sub-area with the relative central points, maybe it works. But it's hard to do it, because GSH have to subdivide a complex area in more little area as regular as possible.

Attachments:

Screen.png, 23 KB

Permalink Reply by Raul Matagringo on October 11, 2012 at 9:13am

yes, looks much better. But how? And then, the perimeter will be not right because you have the fictive "walls" used for cutting down the figure.

The subdivision is also a problem because you have just a bunch of independent lines. You can subdivide a area, or panel wich is giveb by a continuos border (or polyline). Something closed like the ventrices in polygons 0,1,2,3,4,5,0.

One panel could be triangulated, quadrangulated and so on.

Unfortunately you can get the panel first. Thats funny! Your scope is a panel but you need first one...

Permalink Reply by Landscape on October 11, 2012 at 9:32am

mmmm ... duplicate real perimeter, then extend line that could shape regular forms ... could be a right idea?!

Permalink Reply by Raul Matagringo on October 11, 2012 at 9:46am

which line to extend. you dont know the order of your lines in the collection. and what happens when you extend to much (intersection with edge). you also dont know the value for your extension.

see you tomorrow...

Permalink Reply by Danny Boyes on October 11, 2012 at 10:01am