gluUnProject Android OpenGL ES 1.1 Usage

Question

I'm trying to use gluUnProject to convert windows coords to world coords. I'm not trying to get working sample in emulator or older Android systems (with OpenGL ES v1.0) and this is not question about GL function availability. I'm trying to work on the real device with OpenGL ES 1.1 and glGet functions return non-zero results.

There is sample:

public Vector3 translateScreenCoordsToWorld(Vector2 screenCoords) {
gl.glLoadIdentity();

final Matrix4x4 modelViewMatrix = new Matrix4x4();
final Matrix4x4 projectMatrix = new Matrix4x4();

int[] viewVectorParams = new int[4];
gl.glGetIntegerv(GL11.GL_VIEWPORT,viewVectorParams,0);
gl.glGetFloatv(GL11.GL_MODELVIEW_MATRIX,modelViewMatrix.buffer,0);
gl.glGetFloatv(GL11.GL_PROJECTION_MATRIX,projectMatrix.buffer,0);

float[] output = new float[4];
GLU.gluUnProject(
    screenCoords.x, screenCoords.y, 0, 
    modelViewMatrix.buffer, 0, 
    projectMatrix.buffer, 0, 
    viewVectorParams, 0, 
    output, 0);             

return new Vector3(output[0],output[1],output[2]);

}

Matrix4x4 is just a wrapper for float[] buffer;

With this function I'm trying to create a plane to fill all screen or detect maximum world coords for current projection matrix, but it's not working at all because I not really sure that I'm using these functions correctly.

For example when I'm trying run translateScreenCoordsToWorld(new Vector2(480,800)) it returns very small coordinates values Vector3(0.27f, 0.42f, -1.0f).

Could anyone provide good sample usage gluUnProject for GL_PROJECTION mode with one positioned camera.

Updated Thanks for good links. But it' still not working for me :( Now my function looks like:

public Vector3 translateScreenCoordsToWorld(Vector2 screenCoords) {

float winX = screenCoords.x, winY = screenCoords.y, winZ = 0;
winY = (float)currentViewVectorParams[3] - winY;

float[] output = new float[4];
GLU.gluUnProject(
    winX, winY, winZ, 
    currentModelViewMatrix.buffer, 0, 
    currentProjectMatrix.buffer, 0, 
    currentViewVectorParams, 0, 
    output, 0
);              

Vector3 nearPlane = new Vector3(output[0],output[1],output[2]);
winZ = 1.0f;        
GLU.gluUnProject(
    winX, winY, winZ, 
    currentModelViewMatrix.buffer, 0, 
    currentProjectMatrix.buffer, 0, 
    currentViewVectorParams, 0, 
    output, 0
);              
Vector3 farPlane = new Vector3(output[0],output[1],output[2]);

farPlane.sub(nearPlane);
farPlane.div( nearPlane.length());

float dot1, dot2;

Vector3 pointInPlane = new Vector3(), pointPlaneNormal = new Vector3(0,0,-1);
pointInPlane.sub( nearPlane );

dot1 = (pointPlaneNormal.x * pointInPlane.x) + (pointPlaneNormal.y * pointInPlane.y) + (pointPlaneNormal.z * pointInPlane.z);
dot2 = (pointPlaneNormal.x * farPlane.x) + (pointPlaneNormal.y * farPlane.y ) +(pointPlaneNormal.z * farPlane.z);

float t = dot1/dot2;
farPlane.mul(t);

return farPlane.add(nearPlane);
}

and this is where my camera configuring:

public void updateCamera() {
Camera camera = scene.getCamera();
GLU.gluLookAt(gl, 
    camera.position.x, camera.position.y, camera.position.z,
    camera.target.x, camera.target.y, camera.target.z,
    camera.upAxis.x, camera.upAxis.y, camera.upAxis.z
);

    gl.glGetIntegerv(GL11.GL_VIEWPORT,currentViewVectorParams,0);       
    gl.glGetFloatv(GL11.GL_MODELVIEW_MATRIX, currentModelViewMatrix.buffer,0);
    gl.glGetFloatv(GL11.GL_PROJECTION_MATRIX,currentProjectMatrix.buffer,0);
}

The camera configured with the following coords:

camera.position = { 0, 0, 65 };     
camera.target = { 0, 0, 0 }
camera.upAxis = { 0, 1, 0 }

score 3 · Accepted Answer · answered Sep 16 '11 at 14:03

3

Okay, normally when using gluUnProject you are trying to get the world space coordinate of a pixel on screen. You need three pieces of information,(other than the matrices) to put into glUnProject, a screen x coordinate , a screen y coordinate, and the depth buffer value corresponding to that pixel. The process normally goes like this:

Draw a frame, so all the depth buffer information is ready.
Grab viewport/matrices/screen coords
Invert the screen y coordinate.
Read the depth value at a given pixel. This is effectively a normalised z coordinate.
Input the x,y and z into glUnProject.

This ends up giving the world space coordinate of the given fragment of a pixel. A much more detailed guide can be found here. Anyway, I can see two possible mistakes in your code. The first is that normally you have to invert the y screen coord,(the tutorial on the link details why), the second is that you are always passing in 0 as the z value to gluUnProject. Doing this un-projects the vertex as if it was on the near plane. Is this what you want?

Anyway, forgive me if I misread your question.

answered Sep 16 '11 at 14:03

Darcy Rayner

3,385
1
23
15

Thanks! It's good NeHe article. But there is no GL_DEPTH_COMPONENT support in OpenGL ES (Android). I can't call glReadPixels(winX, winY, 1, 1, GL_DEPTH_COMPONENT, GL_FLOAT, &winZ); :( Is there any workaround solution? – abdolence Sep 16 '11 at 20:15
3

You are right, I probably should have remembered that. There are two work arounds, the first is to create a ray between the near and the far plane and run a manual intersection test,(which might be quite tricky), [here](http://softwareprodigy.blogspot.com/2009/08/gluunproject-for-iphone-opengl-es.html), the second is to create a frame buffer object, and read the depth buffer values out from that. Alternatively you could skip gluUnProject and try implementing picking. [See Here](http://stackoverflow.com/questions/7364693/how-to-get-object-in-webgl-3d-space-from-a-mouse-click-coordinate) – Darcy Rayner Sep 16 '11 at 23:05
Interested links again. Thanks! I tried to implement manual intersect, but It's still not working :( Please look at the update – abdolence Sep 23 '11 at 17:51

gluUnProject Android OpenGL ES 1.1 Usage

1 Answers1

Linked