Projecting mesh from single point to geometry

[nkyms]

Hello,

Houdini beginner here, I am looking for a solution to project geometry from a given single point (labelled as the green dot in the diagram, just above the mid point of the sphere) to y=0 to make a semisphere.

The point is to have a flat ground without deformations and changes to the UV projection while a camera is positioned at the given point. So say you had a panoramamic image/video applied to the semisphere, it would look undistorted when viewing from this given point.

 

Thanks in advance!

Edited December 12, 2020 by nkyms

[animatrix]

Hi,

You mean like this?

Then you can just use this inside a Point Wrangle SOP:

if ( @P.y < 0 )
    @P.y = 0;

if ( @P.y < 0 )
    @P.y = 0;

 

[nkyms]

Thanks for the reply.

No, the shape is close to what is intended but what I'd like to do is bring the points in towards the middle.

In theory I need each point below y=0 to move towards 0,0,0 (or a given point) until it hits the y axis.

14 minutes ago, animatrix said:

Hi,

You mean like this?

Then you can just use this inside a Point Wrangle SOP:

if ( @P.y < 0 )
    @P.y = 0;

if ( @P.y < 0 )
    @P.y = 0;

 

 

[animatrix]

Like this?

if ( @P.y < 0 )
    @P = 0;

 

[nkyms]

No, let me reword this. 

I'd like to move each point below the y-axis towards the middle until it reaches the y axis.

Meaning the points will not come to a single point. And when viewing from the middle point, each point will still be at the same angle.

I'd use the ray node, but this only allows for a vector direction.

Edited December 12, 2020 by nkyms

[Librarian]

You can Use Clip Sop with Stereo Projection ON any Given Point Or Geo.Have Fun ..You Have Those Methods Here ON OdFroce Investigate
Link

[nkyms]

I am aware of Clip Sop however I don't recall there being a Stereo Projection functionality. Could you elaborate on this method? 

I've had a search around on the forums, but I still can't find a solution.

Thanks for the help.

[animatrix]

[nkyms]

27 minutes ago, animatrix said:

Thanks for the help so far.

By bringing the points towards the middle of the initial sphere they should hit the bottom of the semisphere along the x,z plane, sorry I think I confused you by saying y-axis. 

The first image you sent is the closest, but the angles of each point viewed from the center are changed. I need each point to be the same angle and projected to the lower base of the semi-circle.

[animatrix]

I see yes this can be done using Ray Plane intersection analytically without creating a plane geometry.

I will post an example when i get home.

 

[Librarian]

v@rayDir = {0, -1, 0};

 

Ray.hiplc

[konstantin magnus]

Inside view / Outside view:
 

hemisphere.hipnc

[konstantin magnus]

Same result along with greetings from stack overflow:

string cam = chs('camera');

matrix xform_cam = optransform(cam);
vector pos_cam = cracktransform(0,0,0,0,0, xform_cam);
vector pos_plane = {0,0,0};
vector nml_plane = {0,1,0};

vector dir_line = normalize(v@P - pos_cam);
float scale = dot(nml_plane, pos_plane) - dot(nml_plane, pos_cam) / dot(nml_plane, dir_line);

v@P = pos_cam + dir_line * scale;

 

hemisphere_calc.hipnc

[Noobini]

i don't know if the effect is same as:

- Point SOP

- preset Flatten Bottom

- adjust to whatever level you like (max half way), doesn't have to be exact hemisphere, animatable.

[nkyms]

14 hours ago, konstantin magnus said:

Same result along with greetings from stack overflow:

string cam = chs('camera');

matrix xform_cam = optransform(cam);
vector pos_cam = cracktransform(0,0,0,0,0, xform_cam);
vector pos_plane = {0,0,0};
vector nml_plane = {0,1,0};

vector dir_line = normalize(v@P - pos_cam);
float scale = dot(nml_plane, pos_plane) - dot(nml_plane, pos_cam) / dot(nml_plane, dir_line);

v@P = pos_cam + dir_line * scale;

 

hemisphere_calc.hipnc

Thanks for the solution.

This did the trick. I still don't really understand what is happening fully in the expression, but I'll have a shot.

[animatrix]

You can also use something like this which is similar to Konstantin's approach:

vector linePlaneIntersection ( vector raypos, raydir, planepos, planedir )
{
    vector diff = raypos - planepos;
    vector d = dot ( diff, planedir ) / dot ( raydir, planedir );
    return raypos - raydir * d;
}

if ( @P.y < 0 )
{
    vector n = normalize ( chv("p") - @P );
    vector p = linePlaneIntersection ( @P, n, 0, { 0, 1, 0 } );
    @P = p;
}

Depending on the P parameter, the result will change, so make sure to set it to non 0, 0, 0. Otherwise all bottom points will intersect at 0, 0, 0. Here I use 0, 0.3, 0:

Edited December 13, 2020 by animatrix

[nkyms]

10 hours ago, animatrix said:

You can also use something like this which is similar to Konstantin's approach:

vector linePlaneIntersection ( vector raypos, raydir, planepos, planedir )
{
    vector diff = raypos - planepos;
    vector d = dot ( diff, planedir ) / dot ( raydir, planedir );
    return raypos - raydir * d;
}

if ( @P.y < 0 )
{
    vector n = normalize ( chv("p") - @P );
    vector p = linePlaneIntersection ( @P, n, 0, { 0, 1, 0 } );
    @P = p;
}

Depending on the P parameter, the result will change, so make sure to set it to non 0, 0, 0. Otherwise all bottom points will intersect at 0, 0, 0. Here I use 0, 0.3, 0:

Thanks, this works well. I'll have a look at your solution too.

Edited December 13, 2020 by nkyms