Translation

3D Transformations

CS 4620 Lecture 3

Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 1 Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 2

Scaling Rotation about z axis

Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 3 Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 4
 Rotation about x axis Rotation about y axis

Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 5 Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 6

Transformations in OpenGL Transformations in OpenGL!

• Stack-based manipulation of model-view transformation, M glMatrixMode(GL_MODELVIEW);
glLoadIdentity();
• glMatrixMode(GL_MODELVIEW) Specifies model-view matrix
{// Draw something:
glPushMatrix();
• glLoadIdentity() M ! 4x4 identity glTranslatef(...);
glRotatef(15f, ...);
• glTranslatef(float ux, float uy, float uz) M ! MT {// set color and draw simplices
glBegin(GL_TRIANGLES);
glColor3f(...);
• glRotatef(float theta, float ux, float uy, float uz) M!MR glVertex3f(...);
glVertex3f(...);
• glScalef(float sx, float sy, float sz) M!MS glVertex3f(...);
glEnd();
}
• glLoadMatrixf(float[] A) M !A (Note: column major) glPopMatrix(); // toss old transform
}
• glMultMatrixf(float[] A) M ! M A (Note: column major) {// Draw something else:
glPushMatrix();
• Manipulate matrix stack using: ...
glPopMatrix(); // toss old transform
– glPushMatrix()
}
– glPopMatrix()
Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 7 Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 8
 Transformations in OpenGL! General Rotation Matrices
• A rotation in 2D is around a point
– Tutors demo
• A rotation in 3D is around an axis
– so 3D rotation is w.r.t a line, not just a point
– there are many more 3D rotations than 2D
• a 3D space around a given point, not just 1D

2D 3D
Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 9 Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 10

Properties of Rotation Matrices Specifying rotations

• Columns of R are mutually orthonormal: RRT=RTR=I • In 2D, a rotation just has an angle
– if it’s about a particular center, it’s a point and angle
• Right-handed coordinate systems: det(R)=1
– Recall definition of det(R)=r1T(r2xr3) • In 3D, specifying a rotation is more complex
– basic rotation about origin: unit vector (axis) and angle
• Such 3x3 rotation matrices belong to group, SO(3) • convention: positive rotation is CCW when vector is pointing at you
– Special orthogonal – about different center: point (center), unit vector, and angle
– Special --> det(R)=1 • this is redundant: think of a second point on the same axis...

• Alternative: Euler angles

– stack up three coord axis rotations
• ZYX case: Rz(az)*Ry(ay)*Rx(ax)
– degeneracies exist for some angles
– E.g., gimbal lock
– Black board
Unlocked Gimbal lock
Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 11 Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 12
 Coming up with the matrix Building general rotations
• Showed matrices for coordinate axis rotations • Using elementary transforms you need three
– but what if we want rotation about some random axis? – translate axis to pass through origin
– rotate about y to get into x-y plane
• Can compute by composing elementary transforms – rotate about z to align with x axis
– transform rotation axis to align with x axis
– apply rotation • Alternative: construct frame and change coordinates
– inverse transform back into position – choose p, u, v, w to be orthonormal frame with p and u matching the
rotation axis
• Just as in 2D this can be interpreted as a similarity transform
– apply similarity transform T = F Rx(! ) F–1

Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 13 Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 14

Orthonormal frames in 3D! Building 3D frames

• Useful tools for constructing transformations • Given a vector a and a secondary vector b
– The u axis should be parallel to a; the u–v plane should contain b
• Recall rigid motions • u = u / ||u||
– affine transforms with pure rotation • w = u x b; w = w / ||w||
– columns (and rows) form right-handed ONB • v=wxu
• that is, an orthonormal basis
• Given just a vector a
– The u axis should be parallel to a; don’t care about orientation about
that axis
• Same process but choose arbitrary b first
• Good choice is not near a: e.g. set smallest entry to 1

Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 15 Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 16
 Building general rotations Derivation of General Rotation Matrix
• Alternative: construct frame and change coordinates • General 3x3 3D rotation matrix
– choose p, u, v, w to be orthonormal frame with p and u matching the
rotation axis • General 4x4 rotation about an arbitrary point
– apply similarity transform T = F Rx(! ) F–1
– interpretation: move to x axis, rotate, move back
– interpretation: rewrite u-axis rotation in new coordinates
– (each is equally valid)

• Or just derive the formula once, and reuse it (more later)

Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 17 Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 18

Building transforms from points Transforming normal vectors

• Recall: 2D affine transformation has 6 degrees of freedom • Transforming surface normals
(DOFs) – differences of points (and therefore tangents) transform OK
– this is the number of “knobs” we have to set to define one – normals do not --> use inverse transpose matrix
• Therefore 6 constraints suffice to define the transformation
– handy kind of constraint: point p maps to point q (2 constraints at once)
– three point constraints add up to constrain all 6 DOFs
(i.e. can map any triangle to any other triangle)
• 3D affine transformation has 12 degrees of freedom
– count them by looking at the matrix entries we’re allowed to change
• Therefore 12 constraints suffice to define the transformation
– in 3D, this is 4 point constraints
(i.e. can map any tetrahedron to any other tetrahedron)

Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 19 Cornell CS4620 Fall 2010 •!Lecture 3 © 2010 Doug James • 20