point normals pointing away

[csp]

Hi all,

I have a group of points from a grid geometry and I want as they are bend to have their normals pointing away perpendicular to their points of the neighbour rows. I can't use edge_dir because there are no edges just points. I have attached two images, the first is the points as result of the pending and the second is just a cylinder but demonstrate has the normals I want them to be.

I have also attached a hip file.

point_normals.hip

[sadhu]

Is this is what you want?

point_normals_soln.hipnc

[csp]

Is this is what you want?

Thanks for your reply sadhu. Yes this is end result I am after but in the original scene the bending has to happen on the points, there is no surface. that's why I had the nodes in this order.

[csp]

I found one way to fix it but is very messy, I have the same process with the bending on the points and repeated on the original surface and then I transfer the normal attribute from the second to the first with an attribtrasferSOP. They dont have for both groups to have the same points. Is there any other more clean and efficient way?

Edited November 22, 2012 by cparliaros

[sadhu]

If you have access to bended surface then attrib transfer is the simplest soln. Another soln attached

point_normals_Vopsoln.hipnc

Edited November 22, 2012 by sadhu

[csp]

If you have access to bended surface then attrib transfer is the simplest soln. Another soln attached

thanks that works great.

EDIT: I just realized when I scaled up the normals that they are not right in your approach because they are pointing away from the center of the group of points not from each point. I guess I have to go back to attribute transfer solution. cheers.

Edited November 22, 2012 by cparliaros

[magneto]

If I understand you correctly, just subtract the center from the point positions in a point SOP like:

NX = $TX - $CEX

etc

After that you have to normalize them so you can do it in a VOPSOP, or just do the whole thing VOPs, using the Bounding Box VOP to calculate the center and then do the same. That's probably what I would prefer

[csp]

well not exactly, this way you have the same result as sadhu latest example where all the normals point away for the center of the object. I have attached a scene with the two ways to attemot fix this. the first named "magneto" I have what you described in the other named "cparliaros" I have my messy fix which gives the correct results.

point_normals_0002.hipnc

Edited November 22, 2012 by cparliaros

[magneto]

I see what you wanted. If the points are ordered, it's easy to do.

All you need to do is in VOPs (pseudo-code):

d = (importattib (ptnum+1) "P") - (importattib ptnum "P")
N = normalize(cross(d, {1,0,0)))
[/CODE]

For the last point, since it doesn't follow the flow of points in one direction, you could copy and mirror it from the first one.

Edited November 22, 2012 by magneto

[junaid]

yes I would do everything in VOP's, it will be bit faster and you could easily add more advance functionality.

some thing like this in attachment.

[junaid]

here is the attachment

point_normals_0003.hipnc

[csp]

here is the attachment

sorry man but you are getting the same results as sadhu last example.

I see what you wanted. If the points are ordered, it's easy to do.

All you need to do is in VOPs (pseudo-code):

d = (importattib (ptnum+1) "P") - (importattib ptnum "P")
N = normalize(cross(d, {1,0,0)))
[/CODE]

For the last point, since it doesn't follow the flow of points in one direction, you could copy and mirror it from the first one.

its really close to what I am getting with the attribtrasfer but they are not that accurate besides the last point. I will probably do it in vopsop using pointcloud which actually simulates the attribtransferSOP but is faster.

thanks for all the replies.

[junaid]

its quite confusing what you want, but whatever works for you !

[sadhu]

This will work.

Separate component of the vectors, set x to zero and subtract y from y and z from z component.

point_normals_FIXED.hipnc

[csp]

This will work.

Separate component of the vectors, set x to zero and subtract y from y and z from z component.

well I think is my mistake that I used this example with that amount of bending. With your fix and junaid's fix, if your change the strength in the twistSOP in something less or more, you will see that the normals will not look right any more. The center point is useful when the object is rounded.

cheers guys

[junaid]

like this ?

[junaid]

like this?

point_normals_0004.hipnc

[edward]

?