Hi I am working on a AR android app. I am using ARToolkit6. In this app I want to view my 3D object( A Cube) on left half of the screen. With this eventually I want to display 3 cubes on the screen each on 1/3 of the screen area.
I was able to scale the 3D object by tweaking ModelView Matrix. What I read so far, I think I need to tweak projection matrix to achieve my goal. I tried looking solutions online. But Couldn't get it to work. Can anyone direct me to right path?
for (int trackableUID : trackableUIDs) {
// If the marker is visible, apply its transformation, and render a cube
if (ARToolKit.getInstance().queryMarkerVisible(trackableUID)) {
float[] projectionMatrix = ARToolKit.getInstance().getProjectionMatrix();
float[] modelViewMatrix = ARToolKit.getInstance().queryMarkerTransformation(trackableUID);
float[] scalingMat = {1, 0, 0, 0, 0, 3.0f, 0, 0, 0, 0, 1.0f, 0, 0.0f, 0, 0, 1};
float[] newModelView = modelViewMatrix;
multiplyMM(newModelView, 0, modelViewMatrix, 0, scalingMat, 0);
cube.draw(projectionMatrix, newModelView);
}
I followed the this link Set origin to top-left corner of screen in OpenGL ES 2 and (OpenGL ES) Objects away from view centre are stretched. So I translated the modelView Matrix but it doesn't solve the problem, the 3D object appears at the center of the screen. Can you explain how should I approach this problem? Thanks
#Override
public void draw() {
super.draw();
GLES20.glEnable(GLES20.GL_CULL_FACE);
GLES20.glEnable(GLES20.GL_DEPTH_TEST);
GLES20.glFrontFace(GLES20.GL_CCW);
// Look for trackables, and draw on each found one.
for (int trackableUID : trackableUIDs) {
// If the marker is visible, apply its transformation, and render a cube
if (ARToolKit.getInstance().queryMarkerVisible(trackableUID)) {
float[] projectionMatrix = ARToolKit.getInstance().getProjectionMatrix();
float[] modelViewMatrix = ARToolKit.getInstance().queryMarkerTransformation(trackableUID);
float[] scalingMat = {1, 0, 0, 0, 0, 3.0f, 0, 0, 0, 0, 1.0f, 0, 0.0f, 0, 0, 1};
multiplyMM(modelViewMatrix, 0, scalingMat, 0, modelViewMatrix, 0);
float[] rightModelMatrix = new float[16];
Matrix.setIdentityM(rightModelMatrix, 0);
// Translate outer sphere by 5 in x.
Matrix.translateM(rightModelMatrix, 0, 5.0f, 0.0f, 0.0f);
Matrix.multiplyMM(modelViewMatrix, 0, rightModelMatrix, 0, modelViewMatrix, 0);
cube.draw(projectionMatrix, modelViewMatrix);
}
}
Also tried this but the object gets displayed at the center of the screen.
glMatrixMode(GL_PROJECTION);
glTranslatef(5f, 0f, 0f);
Related
I am having some problems with proper scene manipulations with OpenGL ES 2.0. On my screen I want to draw a rectangle and a cube. Rectangle should be moved to the bottom and scaled while cube should be moved to the top and scaled. What's more, I want to rotate this cube in all 3 axis.
This is what I've created so far:
#Override
public void onDrawFrame(GL10 gl) {
float[] scratch = new float[16];
GLES20.glClear(GLES20.GL_COLOR_BUFFER_BIT);
Matrix.setLookAtM(viewMatrix, 0, 0, 0, -6, 0f, 0f, 0f, 0f, 1.0f, 0.0f);
Matrix.multiplyMM(MVPMatrix, 0, projectionMatrix, 0, viewMatrix, 0);
Matrix.translateM(scratchMatrix1, 0, 0, -1.1f, 0);
Matrix.scaleM(scratchMatrix3, 0, 0.5f, 0.5f, 0.5f);
Matrix.multiplyMM(scratchMatrix2, 0, scratchMatrix1, 0, scratchMatrix3, 0);
Matrix.multiplyMM(scratch, 0, scratchMatrix2, 0, MVPMatrix, 0);
rectangle.draw(scratch);
Matrix.setRotateM(scratchMatrix1, 0, angleXVal, 1.0f, 0, 0);
Matrix.setRotateM(rotationMatrix, 0, angleYVal, 0, 1.0f, 0);
Matrix.multiplyMM(scratchMatrix2, 0, rotationMatrix, 0, scratchMatrix1, 0);
Matrix.setRotateM(rotationMatrix, 0, angleZVal, 0, 0, 1.0f);
Matrix.multiplyMM(scratchMatrix3, 0, rotationMatrix, 0, scratchMatrix2, 0);
Matrix.translateM(scratchMatrix1, 0, 0, 0.3f, 0);
Matrix.multiplyMM(scratchMatrix2, 0, scratchMatrix3, 0, scratchMatrix1, 0);
Matrix.scaleM(scratchMatrix1, 0, 0.8f, 0.8f, 0.8f);
Matrix.multiplyMM(scratchMatrix3, 0, scratchMatrix2, 0, scratchMatrix1, 0);
Matrix.multiplyMM(scratch, 0, scratchMatrix2, 0, MVPMatrix, 0);
cube.draw(scratch);
}
The problem is that when I am rotating the cube also the rectangle is moving. It is this same with scale and translating of cube - it is pulling the rectangle.
For me it looks like the matrices are connected somewhere but I don't see where, because the rectangle is drawn first and scratch matrix is overwritten later.
Assuming that is your complete onDrawFrame() code, scratchMatrix1 etc are class member variables that persist from frame to frame.
That means that translations from one frame will carry over to the next, because Matrix.translateM() applies a translation to an existing matrix transform by adding to it, rather than setting the matrix to a given translation matrix and overwriting the previous values. So the translations will accumulate over several frames.
That means that when you translate the cube, the translation will still be sitting in scratchMatrix1 the next time onDrawFrame is called, and the rectangle translation will be added to it rather than be set afresh.
The same goes for Matrix.scaleM().
You can rectify this by initializing all the scratch matrices to identity at the start of the function, and again before beginning the calculation of the cube transform matrix:
Matrix.setIdentityM(scratchMatrix1, 0);
Matrix.setIdentityM(scratchMatrix2, 0);
Matrix.setIdentityM(scratchMatrix3, 0);
Matrix.setIdentityM(scratch, 0);
I want to rotate an element around a specific point defined by me and dynamically changed.
I am orientating myself at the guidelines from the google developers site.
My first approach is this:
scratch = new float[16];
Matrix.setIdentityM(mRotationMatrix, 0);
Matrix.setRotateM(mRotationMatrix, 0, angle, 0, 0, 1f);
Matrix.multiplyMM(scratch, 0, mMVPMatrix, 0, mRotationMatrix, 0);
element.draw(scratch);
This rotates the object around the center of the screen.
What do I have to add/change to make the object rotate around some other point?
Add a translation operation.
Grafika's Sprite2d class provides an example:
/**
* Re-computes mModelViewMatrix, based on the current values for rotation, scale, and
* translation.
*/
private void recomputeMatrix() {
float[] modelView = mModelViewMatrix;
Matrix.setIdentityM(modelView, 0);
Matrix.translateM(modelView, 0, mPosX, mPosY, 0.0f);
if (mAngle != 0.0f) {
Matrix.rotateM(modelView, 0, mAngle, 0.0f, 0.0f, 1.0f);
}
Matrix.scaleM(modelView, 0, mScaleX, mScaleY, 1.0f);
mMatrixReady = true;
}
This positions the object, then rotates it around the center of the object.
You need to translate the matrix in the reverse direction of the point first, then rotate and then translate it back. Look at it as if the rotation is always rotating around the center of the world, and the translation moves the center of the world.
Something like this (untested):
scratch = new float[16];
Matrix.setIdentityM(mRotationMatrix, 0);
Matrix.translateM(mRotationMatrix, 0, -x, -y, -z);
Matrix.rotateM(mRotationMatrix, 0, angle, 0, 0, 1f);
Matrix.translateM(mRotationMatrix, 0, x, y, z);
Matrix.multiplyMM(scratch, 0, mMVPMatrix, 0, mRotationMatrix, 0);
element.draw(scratch);
The x, y and z values need to be calculated as the delta between the current position of the object in the world and the position of the point you want to rotate around. You need to do that calculation yourself, but that's pretty trivial.
I have, a problem with the setLookAtM function. My goal is to create a cube within a cube something like this (yep, it's paint :P ):
So basically everything works... almoust... I have the smaller cube and I have the bigger one.
However, there is a problem. I created the bigger one with coords from -1 to 1 and now I want to upscale it. With scale 1.0f i have something like this (the inner cube is rotating):
And thats good, but now... when I try to scale the bigger cube (so that it looks like in the paint drawing) the image goes black or white (i guess it's because the "camera" looks at the white cube but still i dont know why does my inner cube disappear :/ I don't understand what I'm doing wrong. Here is my code:
public void onDrawFrame(GL10 unused) {
float[] scratch = new float[16];
GLES20.glClear(GLES20.GL_DEPTH_BUFFER_BIT | GLES20.GL_COLOR_BUFFER_BIT);
GLES20.glEnable(GLES20.GL_DEPTH_TEST);
Matrix.setLookAtM(mViewMatrix, 0, 0, 0, -5.0f, 0f, 0f, -1.0f, 0f, 1.0f, 0.0f);
Matrix.multiplyMM(mMVPMatrix, 0, mProjectionMatrix, 0, mViewMatrix, 0);
mRoom.mScale = 1.0f;
Matrix.setIdentityM(mScaleMatrix, 0);
Matrix.scaleM(mScaleMatrix, 0, mRoom.mScale, mRoom.mScale, mRoom.mScale);
float[] scaleTempMatrix = new float[16];
Matrix.multiplyMM(scaleTempMatrix, 0, mMVPMatrix, 0, mScaleMatrix, 0);
mRoom.draw(scaleTempMatrix);
When I set for example:
mRoom.mScale = 3.0f;
And
Matrix.setLookAtM(mViewMatrix, 0, 0, 0, -2.0f, 0f, 0f, 0.0f, 1.0f, 1.0f, 0.0f);
My camera should be at (0, 0, -2) looking at (0,0, -1) and it should be inside the white cube (since scale is 3.0 so the coords should be from -3 to 3 right?) But all I get is a white screen without the smaller cube rotating inside :/
If your scale is 3x in this code, then your visible coordinate range is actually going to be [-1/3,1/3].
You are thinking about things backwards, it might help if you considered the order in which the scale operation is applied. Right now you are scaling the object-space coordinates, then applying the view matrix and then projection. It may not look that way, but that is how matrix multiplication in GL works; GL effectively flips the operands when it does matrix multiplication and matrix multiplication is not commutative.
I believe this is what you actually want:
public void onDrawFrame(GL10 unused) {
float[] scratch = new float[16];
GLES20.glClear(GLES20.GL_DEPTH_BUFFER_BIT | GLES20.GL_COLOR_BUFFER_BIT);
GLES20.glEnable(GLES20.GL_DEPTH_TEST);
Matrix.setLookAtM(mViewMatrix, 0, 0, 0, -5.0f, 0f, 0f, -1.0f, 0f, 1.0f, 0.0f);
mRoom.mScale = 3.0f;
Matrix.setIdentityM(mScaleMatrix, 0);
Matrix.scaleM(mScaleMatrix, 0, mRoom.mScale, mRoom.mScale, mRoom.mScale);
Matrix.multiplyMM(mMVPMatrix, 0, mScaleMatrix, 0, mProjectionMatrix, 0);
Matrix.multiplyMM(mMVPMatrix, 0, mMVPMatrix, 0, mViewMatrix, 0);
mRoom.draw(mMVPMatrix);
I am using default android sample code
http://developer.android.com/training/graphics/opengl/touch.html
In this sample we can rotate triangle by toucht events.
I want just to add movement by x,y axiss for test purposes.
The point that triangle behaviour is not as i am expecting. What i am doing wrong?
Code from tutorial with my new row hilighted:
public void onDrawFrame(GL10 unused) {
// Draw background color
GLES20.glClear(GLES20.GL_COLOR_BUFFER_BIT);
// Set the camera position (View matrix)
Matrix.setLookAtM(mVMatrix, 0, 0, 0, -3, 0f, 0f, 0f, 0f, 1.0f, 0.0f);
// Calculate the projection and view transformation
Matrix.multiplyMM(mMVPMatrix, 0, mProjMatrix, 0, mVMatrix, 0);
// Draw square
mSquare.draw(mMVPMatrix);
**//Translating this matrix 'brakes' triangle
-> Matrix.translateM(mMVPMatrix, 0, 0, pos, -1.0f);
//NOTHING happens here: ??? Why?
-> Matrix.translateM(mRotationMatrix, 0, pos, 0, -1.0f);**
// Create a rotation for the triangle
// long time = SystemClock.uptimeMillis() % 4000L;
// float angle = 0.090f * ((int) time);
Matrix.setRotateM(mRotationMatrix, 0, mAngle, 0, 0, -1.0f);
// Combine the rotation matrix with the projection and camera view
Matrix.multiplyMM(mMVPMatrix, 0, mRotationMatrix, 0, mMVPMatrix, 0);
// Draw triangle
mTriangle.draw(mMVPMatrix);
}
Default behaviour:
With my code:
Thanks for a icrev comment:
You can not do translation / rotation / scaling on MVP matrix and get results as you expect.
you must translate / rotate your object in model matrix (or in View matrix for camera trans/rotation).
Look at this The purpose of Model View Projection Matrix to understand better what you need to do
These are the steps:
set M matrix to Identity matrix. Translate or rotate it. Be aware of gimbal lock (en.wikipedia.org/wiki/Gimbal_lock)
set V matrix Matrix.setLookAtM(mVMatrix, 0, 0, 0, -3, 0f, 0f, 0f, 0f, 1.0f, 0.0f); 3. you already have projection matrix (in your case mProjMatrix)
multippy M * V * P to recieve final MVP matrix
In Android OpenGL
it has the command setLookAtM to specific the position for the camera view
Matrix.setLookAtM(mViewMatrix, 0, eyeX, eyeY, eyeZ, lookX, lookY,
lookZ, upX, upY, upZ);
If I rotate camera, by using command rotateM
Matrix.rotateM(mViewMatrix, 0 , angle , 0.0f, 1.0f, 0.0f);
Then, How could I get the 'LookAt' the exact camera view from mViewMatrix ?
the camera view position, i want is x,y,z that camera look at
Be careful, as calling rotateM doesn't technically rotate the camera, it applies a rotation to any future objects drawn. (it doesn't mean the same thing).
However in general, if you have a view matrix, and you want to see which direction it is facing, you want to transform the eye space forward vector (0,0,1) by the inverse of the view matrix.
As the view matrix transforms vectors in world space into eye space, the inverse view matrix transforms vectors from eye space into world space.
So you can:
Apply arbitrary operations to the current view matrix
Take its inverse
Multiply the inverse matrix by (0,0,1,0). (note this is the same as just pulling out the third column of the inverse matrix).
After #3 you will have the direction of the camera eye in world space. Add this to the eye's position, and you should know at what point it is pointing.
I have finished my question
and this is my solution
first, I need to made this 3 matrix to get result of x, y, z position
private float[] mRotXMatrix = new float[] {
1, 0, 0, 0,
0, 0, 1, 0,
0, 1, 0, 0,
0, 0, 0, 1 };
private float[] mRotYMatrix = new float[] {
0, 0, 1, 0,
0, 1, 0, 0,
1, 0, 0, 0,
0, 0, 0, 1 };
private float[] mRotZMatrix = new float[] {
0, 1, 0, 0,
1, 0, 0, 0,
0, 0, 1, 0,
0, 0, 0, 1 };
Then in onSensorChanged(SensorEvent e) method I rotate all these 3 matrix together with my camera view like this
Matrix.multiplyMM(mViewMatrix, 0, deltaRotationMatrix, 0, mViewMatrix, 0);
Matrix.multiplyMM(mRotXMatrix, 0, deltaRotationMatrix, 0, mRotXMatrix, 0);
Matrix.multiplyMM(mRotYMatrix, 0, deltaRotationMatrix, 0, mRotYMatrix, 0);
Matrix.multiplyMM(mRotZMatrix, 0, deltaRotationMatrix, 0, mRotZMatrix, 0);
And to get the X, Y, Z of my camera view, just get it from these matrix
viewX = mRotXMatrix[2];
viewY = -mRotYMatrix[6]; // +/- up to your initial camera view
viewZ = mRotZMatrix[10];