read 3d-positions - gluunproject?

Programming Questions

beginner_..

read 3d-positions - gluunproject?

in Programming Questions • 3 years ago

hello

I've got a problem with the 3d-positions. I've read some threads about this topic but it didn't really help me. so now I aks here if you know more..

I'd like to compare the 3d-positions of a point or a rect or also a sphere with the mouseX and mouseY, so that if the mouse goes over an object, a boolean is set true.

It must be a problem of maths but I'm not that good at it so I looked ob the web. Many people had this problem and I read something about gluunproject, but I didn't find some explanations nor a site to download everything..

so my question:

a) is it sensible to use gluunproject (if yes: can you explain me how to make it work?) or

b) is itpossible to work it out on my own or

c) is there another possibility to work it out?

thanks for your answers.. would really help me..

Replies(8)

bjordal

Re: read 3d-positions - gluunproject?

3 years ago

Well your problem is that the mouse doesn't have a Z or it's Z is always 0 (depending on how you look at it). So you could use the dist() method see if the mouse is over a object in 3D space but again you're limited by the fact that the mouse can only "collide" with objects that has a z really close to 0 and therefor this is probably not what you are looking for.

However what you probably want is the check the mouse against a 2D-projection of the 3D objects or as you put it, " if the mouse goes over an object". What you are looking for it this awesome code that i think antiplastik posted on the wiki: http://wiki.processing.org/w/Picking_with_a_color_buffer

beginner_..

Re: Re: read 3d-positions - gluunproject?

3 years ago

would this also work with another z-translation, with smaller and bigger objects and with points in three dimensions?

and there I have to work with a camera. I've got a coordinate system programmed without camera. If I put a camera in it, everything changes, or not? so is it possible to work with a camera without changing everything or is there another possibility?

bjordal

Re: Re: read 3d-positions - gluunproject?

3 years ago

You don't need to use a camera it's just how they do it in that example.

beginner_..

Re: read 3d-positions - gluunproject?

3 years ago

to be honest don't have an idea how this works.. I've tried on my own to change this code but after deleting the "camera"-code, it didn't work anymore..

so what is the essential thing I need, what is not that important?

I'm sorry that I have this stupid questions, but it's not that easy to me..

Jack Kern

Re: read 3d-positions - gluunproject?

3 years ago

Well the way I've always seen 3D picking done is using GLU unproject or similar means of getting the inverse projection matrix. You can use it to get the 3D point on the camera plane, then you create a ray using that point as the starting point, and the direction vector is the difference between that point and the camera position (then unitized/normalized usually).

Then, you use that ray to test intersections with everything visible & pick-able in your scene. You'll probably want things organized in some kind of spatial partition if you have a lot of objects to test intersections with. This keeps things running smoothly.

It's also a really nice way to place objects in the scene, and move objects... because you can get the intersection of the ray and the ground plane and position objects there.

Jack

http://www.jackkern.com
http://www.twitter.com/ElectricJack

Jack Kern

Re: read 3d-positions - gluunproject?

3 years ago

I knew I had a code snippet somewhere:

// ------------------------------------------------------------------------------------------ //
public static void calculateScreenRay( GLU glu, float x, float y, float width, float height
, Vector3 vRayPosOut, Vector3 vRayDirOut )
{
glu.gluUnProject( (double)x, (double)(height - y), (double)0.1, glViewMats.modelview, 0
, glViewMats.projection, 0, glViewMats.viewport, 0, glViewMats.out, 0 );
vRayPosOut.set( (float)glViewMats.out[0], (float)glViewMats.out[1], (float)glViewMats.out[2] );
glu.gluUnProject( (double)x, (double)(height - y), (double)0.9, glViewMats.modelview, 0
, glViewMats.projection, 0, glViewMats.viewport, 0, glViewMats.out, 0 );
vRayDirOut.set( (float)glViewMats.out[0], (float)glViewMats.out[1], (float)glViewMats.out[2] );
vRayDirOut.dec( vRayPosOut );
vRayDirOut.nrmeq();
}
// ------------------------------------------------------------------------------------------ //
public static Vector3 calculateIntersect( GLU glu, float x, float y, float width, float height
, float[] plane )
{
Vector3 vWorld = new Vector3();
Vector3 vRayPos = new Vector3();
Vector3 vRayDir = new Vector3();
calculateScreenRay( glu, x, y, width, height, vRayPos, vRayDir );
// Finally set the mouse world position to the intersection location
float[] result = rayPlaneIntersect( vRayPos, vRayDir, plane );
if( result != null ) vWorld.set( result );
return vWorld;
}
// ------------------------------------------------------------------------------------------ //
public static float[] rayPlaneIntersect( Vector3 vRayPos, Vector3 vRayDir, float[] plane )
{
Vector3 planeNormal = new Vector3( plane );
float denom = vRayDir.dot( planeNormal ); // Calculate the denominator first
if( Math.abs(denom) < 0.001f ) return null; // Make sure we won't create a divide by zero
float numer = vRayPos.dot( planeNormal ) + plane[3]; // Calculate the numerator next
float time = -numer / denom; // Get the time of intersection
return vRayDir.mul( time ).inc( vRayPos ).toFloats(); // Return the point of intersection
}

These rely on my Vector3 class that's part of my caffeine library, but I will leave it as an exercise for the user to port it to PVector if you don't want to install caffeine just for Vector3 =)

This still leaves the picking up to you, but ray/sphere intersection tests are a good way to start, then you can test polygon intersection by calculating the plane for each polygon/triangle and checking for simple 2D containment of the intersection point and the polygon on the plane.

Jack

[edit]

You'll also need these...

// ------------------------------------------------------------------------------------------ //
public static void updateMatrices( GL gl, GfxViewMatrices mats )
{
gl.glGetDoublev ( gl.GL_MODELVIEW_MATRIX, mats.modelview, 0 );
gl.glGetDoublev ( gl.GL_PROJECTION_MATRIX, mats.projection, 0 );
gl.glGetIntegerv ( gl.GL_VIEWPORT, mats.viewport, 0 );
}
public class GfxViewMatrices
{
public double [] modelview = new double [16];
public double [] projection = new double [16];
public int [] viewport = new int [4];
public double [] out = new double [4];
}

http://www.jackkern.com
http://www.twitter.com/ElectricJack

beginner_..

Re: Re: read 3d-positions - gluunproject?

3 years ago

what you're explaining seems to be exactly what I'm looking for. I also thought of an inverse matrix. BUt I'm not that good at maths so I wasn't able to do it on my own..

What you've posted , is this the code I can only put in a sketch and should work? ( I didn't yet have the time to have a deeper look on it) but there's an error: "expecting class, found void"...

so what's my fault?

thanks for your help by the way.. it's exactly what I'm looking for..

beginner_..

Re: read 3d-positions - gluunproject?

3 years ago

could someone tell me why this error occured?

thanks..

Top Reply