Boundingbox In World Space

[yunjun]

Hey, guys,

I want to create a bounding box for all selected objects in my scene and actually got help from an earlier post at: http://www.highend3d.com/boards/index.php?...09&hl=space.

The post explained very well the routine to get MBoundingBox in the world space.

so I wrote the following:

CODE
    MSelectionList selection;
    MGlobal::getActiveSelectionList( selection );

    MDagPath dagPath;
    MObject component;

    std::vector vmbb = std::vector < MBoundingBox >();

    MItSelectionList iter( selection, MFn::kTransform);
    for (; !iter.isDone(); iter.next() )
    {
        stat=iter.getDagPath( dagPath, component );
        if( stat )
        {
            MObject moTransform = dagPath.transform();
            getBoundingBoxesAsWorld(moTransform, dagPath,vmbb);
        }
    }

    for (size_t i = 1; i < vmbb.size(); i++)
        vmbb[0].expand(vmbb[i]);

    MPoint bCenter, bMax, bMin;
    bCenter = vmbb[0].center();
    bMax = vmbb[0].max();
    bMin = vmbb[0].min();

    then do my own things......

the getBoundingBoxesAsWorld function is copied from the code by Daniel_Faust on this earlier post as below (with a minor change to show where I have problem):
CODE
void getBoundingBoxesAsWorld(MObject &moTransform, const MDagPath &mdpBase, std::vector &vmbb)
{
    MFnTransform mftTransform(moTransform);
    unsigned int iChildCount =mftTransform.childCount();
    for (unsigned int iChildIndex = 0; iChildIndex<iChildCount; iChildIndex++)
    {
        MObject moChild = mftTransform.child(iChildIndex);
        if (moChild != mdpBase.node() )
        {
            if ( moChild.apiType() == MFn::kTransform || moChild.apiType()==MFn::kPluginTransformNode )
            {
                getBoundingBoxesAsWorld(moChild,mdpBase,vmbb);
            }
            else
            {
                MFnDagNode mfdnChild(moChild);
                MBoundingBox mbbChild = mfdnChild.boundingBox();
                MDagPath mdpChild;
                mfdnChild.getPath(mdpChild);
                MMatrix m1 = mdpBase.inclusiveMatrixInverse();
                MMatrix m2 = mdpChild.inclusiveMatrix();
                MMatrix mmiChild = m2*m1                    -----------**----------
                mbbChild.transformUsing(mmiChild);
                vmbb.push_back(mbbChild);
            }
        }
    }
}

but it seemed that the returned MBoundingBoxes are not always correct. for example, at the line I marked as ----*----, many times, m2m1 = identity matrix, so the bounding box end up still being placed at the origin rather than where it should be.

Anyone has some idea? Thank you!

[yunjun]

I guess I have a more general question to ask:

How to transform an object from object space to the world space and vice versa? which class is the most convenient to use? e.g. MFnDagNode?

[yunjun]

Another interesting observation. I am not sure if it is a bug of my code or it is what it should be.

say I have a bounding box (cube) with the length of side as 1 and centering at origin. If I rotate this bounding box (suppose it to be bbox1) 45 degrees along an axis, say x-axis, then in the world space, the bounding box of bbox1 will be bigger, not having all sides as 1, but with the sides in y-z plane being 1.41. that means the bounding box expands a bit.

So if I first retrieve the bounding boxes one by one of all objects in my scene, transform them into world space and do the expand() to get a final bounding box. It is bigger than it should be.
Could this be the reason why my final bounding box looks very loose (with some margin), especially when there are rotations in the transform?

Thank you.

[yunjun]

Hey, Guys,

I found an easier way to get the bounding box for the whole scene.

CODE
    std::vector vmbb = std::vector < MBoundingBox >();
    MDagPath dagPath;
    MItSelectionList iter( selection, MFn::kTransform);
    for (; !iter.isDone(); iter.next() )
    {
        stat=iter.getDagPath( dagPath, component );

        if( stat )
        {
            ////extract boundingbox
            MFnDagNode dagNodeFn(dagPath, &stat);
            MBoundingBox b = dagNodeFn.boundingBox(&stat);
            MMatrix m = dagPath.exclusiveMatrix();
            b.transformUsing(m);
            vmbb.push_back(b);
        }
    }

    for (size_t i = 1; i < vmbb.size(); i++)
    {
        vmbb[0].expand(vmbb[i]);
    }

Earlier I was using the MDagPath.inclusiveMatrix, the matrix includes all transforms in the path (including the the end Node in the Path if it is a Transform!!). Thus, because the MFnDagNode.boundingBox already includes the last transformation. the MboundingBox.transformUsing(matrix) applies the same matrix for a second time and causes the error. Instead using the exclusiveMatrix() solves the problem (only based on my limited observations, don't know what will happen when the scene becomes more complicated.)