[Ardor3d] Exact bounding boxes

Hi,

is there a way to determine the exact bounding box of a node in Ardor3d? The provided "getWorldBounds" method does not return the expected result.

Imagine you have a sphere with radius 0.5. If you compute the axis-aligned world bounding box you get size 1 in all three dimensions.

Now, imagine you have the same sphere and rotate it around x by 45 degrees. You would expect the same bounding box but "getWorldBounds" returns a different result. It seems like there is a invisible box around the sphere which gets rotated and the bounds are computed for that box.

Is there a way to avoid this behavior? I couldn't find a solution for it.

This is the code:

BoundingVolume worldBounds = sphereNode.getWorldBound();
BoundingBox worldBoundingBox = (BoundingBox)worldBounds.asType(Type.AABB);

Regards,

Stefan

Hi,

I think I found it. I had to use the OrientedBoundingBox and update its axis after retrieving the world bounds like this:

final OrientedBoundingBox boundingBox = (OrientedBoundingBox)spatial.getWorldBound().asType(Type.OBB);
boundingBox.setXAxis(new Vector3(1, 0, 0));
boundingBox.setYAxis(new Vector3(0, 1, 0));
boundingBox.setZAxis(new Vector3(0, 0, 1));
                       
(BoundingBox)boundingBox.asType(Type.AABB);

Regards,

Stefan

Hi

Are you sure that your bounding box has been updated after rotating it?

Hi Julien,

sorry, I haven't seen your post.

I just realized that the problem still exists. Weird. I was sure that I solved it.

Do you know more about this topic? This problem applies to all kinds of geometry (e.g. meshes).

Regards,

Stefan

Hi

What do you do to indicate to JogAmp's Ardor3D Continuation that it needs to recompute the bounding volumes?

Hi,

I want to compute the extent of the object and draw a box around it.

Regards,

Stefan

I'd like you to tell me whether you call Mesh.updateWorldBound(boolean) or Spatial.updateGeometricState(double, boolean) or Spatial.updateGeometricState(double).

When you rotate a Spatial, the engine updates both the transform and the bounding.

You can use setModelBound with autoCompute too.

The extent of the object is stored into the bounding box.

Moreover, use a BoundingSphere as a bounding volume of a sphere instead of tinkering a BoundingBox. Call getRadius() on it, use that to draw an orthogonal box around the sphere.

This would work, but imagine the sphere is defined in a geometry file as a pile of triangles that is not known to be a sphere (no primitive). Then it breaks again.

I recognized something: at the time when the code was working correctly, the OrientedBoundingBox's internal attributes "_xAxis", "_yAxis", and "_zAxis" were different from (1,0,0) (0,1,0) (0,0,1) - according to the rotation.

The code

boundingBox.setXAxis(new Vector3(1, 0, 0));
boundingBox.setYAxis(new Vector3(0, 1, 0));
boundingBox.setZAxis(new Vector3(0, 0, 1));

fixed this issue and the bounding box was axis-aligned again. Everything was fine.

At the moment, the internal variables "_xAxis", "_yAxis", and "_zAxis" still have default direction (1,0,0) (0,1,0) (0,0,1) after the call to "spatial.getWorldBound().asType(Type.OBB);".

It seems like the rotation has no influence on the axis of the OrientedBoundingBox anymore - no idea why. I am sure that it had an influence before. Up to now, I could not reproduce the previous behavior.

Regards,

Stefan

Hi,

I think I found it: you have to initialize the bounding box of the sphere using an OrientedBoundingBox like this:

               
OrientedBoundingBox boundingBox = new OrientedBoundingBox();
boundingBox.correctCorners = true;
                       
sphere.setModelBound(boundingBox, true);



The problem was that I had initialized the bounding box of the sphere with an ordinary BoundingBox. Now, it looks good again.

Regards,

Stefan

Actually, I had some difficulties to understand what you expected.

If you call computeFromPoints(), it will be reset to "false".