Hi Roland,
This looks like bug
4428412 - AxisAngle set(Matrix ) fail when rotation is PI about X/Y/Z
- Kelvin
---------
Java 3D Team
Sun Microsystems Inc.
>Delivered-To: [EMAIL PROTECTED]
>X-Sender: [EMAIL PROTECTED]
>Mime-Version: 1.0
>Date: Fri, 21 Sep 2001 12:43:38 +0200
>From: Roland Holm <[EMAIL PROTECTED]>
>Subject: [JAVA3D] Retrieving AxisAngle4d from Transform3D
>To: [EMAIL PROTECTED]
>
>Hi!
>
>I know this topic has been discussed, but I found some weird
>results.
>
>Retrieving an AxisAngle4d from a Transform3D transform should work like this:
>
>
>Quat4d rotQuat = new Quat4d();
>transform.get(rotQuat);
>AxisAngle4d rotation= new AxisAngle4d();
>rotation.set(rotQuat);
>
>or
>
>Quat4d rotQuat = new Quat4d();
>Vector3d anyTranslation = new Vector3d();
>transform.get(rotQuat, anyTranslation);
>AxisAngle4d rotation= new AxisAngle4d();
>rotation.set(rotQuat);
>
>However, under certain circumstances the resulting rotation in the
>AxisAngle4d is a null rotation (0.0 1.0 0.0 0.0), even though there is
>a non-null rotation in the transformation matrix.
>
>Extracting the rotation with the following rough method works correctly:
>
>
>public static AxisAngle4d extractRot(Transform3D trans)
> {
> AxisAngle4d a = new AxisAngle4d();
> Matrix3d r = new Matrix3d();
> trans.get(r);
> // calculate angle
> double calc = (r.m00 + r.m11 + r.m22-1)/2;
> if(calc>1.0) calc = 1.0;
> else if(calc<-1.0) calc = -1.0;
> a.angle = Math.acos(calc);
>
> // No angle equals no rotation, so we are done
> if(a.angle==0.0) return new AxisAngle4d(0,1,0,0);
>
> // calculate axis (note that we have to handle the case of angle==PI
>seperately!)
> double delta= 0.01; // accuracy of the equality check
> if(a.angle<=Math.PI+delta && a.angle>=Math.PI-delta) {
> if(r.m00>=r.m11 && r.m00>=r.m22) {
> a.x = (Math.sqrt(r.m00 - r.m11 - r.m22 +1))/2.0;
> a.y = r.m01 / (2*a.x);
> a.z = r.m02 / (2*a.x);
> } else if(r.m11>=r.m00 && r.m11>=r.m22) {
> a.y = (Math.sqrt(r.m11 - r.m00 - r.m22 +1))/2.0;
> a.x = r.m01 / (2*a.y);
> a.z = r.m12 / (2*a.y);
> } else if(r.m22>=r.m11 && r.m22>=r.m00) {
> a.z = (Math.sqrt(r.m22 - r.m00 - r.m11 +1))/2.0;
> a.x = r.m02 / (2*a.z);
> a.y = r.m12 / (2*a.z);
> }
> } else {
> Vector3d v = new Vector3d(r.m21-r.m12, r.m02-r.m20, r.m10-r.m01);
> double len = v.length();
> Vector3d u = new Vector3d(v.x/len, v.y/len, v.z/len);
> a.x = u.x;
> a.y = u.y;
> a.z = u.z;
> }
> return a;
> }
>
>
>Consider the following transformation matrix:
>
>Matrix4d mat = new Matrix4d(0.9999999999994165, 1.0799814313528522E-6,
>6.437206877160347E-14, -221.0680620796118,
>1.0799814313528522E-6, -1.0000001192084251, -5.83057446579071E-13,
>137.63260064886325,
>-6.437206898336171E-14, 5.830574465790711E-13, -1.0000001192090082,
>379.91844909062524,
>0.0, 0.0, 0.0, 1.0);
>transform.set(mat);
>
>The first way retrieves an AxisAngle4d of
>0.0 1.0 0.0 0.0
>
>while the above method correctly extracts:
>1.0000000298021061 5.399906995835662E-7 3.2186033426590167E-14
>3.141592653589793
>
>
>However, explicitly generating a transformation with
>
>AxisAngle4d aa = new AxisAngle4d(1.0000000298021061, 5.399906995835662E-7,
>3.2186033426590167E-14, 3.141592653589793);
>transform.setRotation(aa);
>
>and retrieving the rotation back with the first method results in correct
>values.
>It's just in these cases like the above transformation matrix.
>
>
>Best regards,
>Roland
>
>===========================================================================
>To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
>of the message "signoff JAVA3D-INTEREST". For general help, send email to
>[EMAIL PROTECTED] and include in the body of the message "help".
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff JAVA3D-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
