I have a bunch of coplanar points, who sit on plane P. I want to translate the plane to Q with the points upon. Here is what I do:

I take three points from P, say a, b, c. Their centroid is denoted by c(a,b,c).

I pick the local origin as O = c(a,b,c) and local XYZ as:

x = normalize(Oa);
v = normalize(Ob);
z = x.crossProduct(v);
y = x.crossProduct(z);

I apply the same method to their global coordinates on Q, which are d, e, f.

O_ = c(d,e,f);
x_ = normalize(Od);
v_ = normalize(Oe);
z_ = x_.crossProduct(v_);
y_ = x_.crossproduct(z_);

Then, I use the transformation matrix as follows:

Matrix input = [x,y,z];
Matrix output = [x_,y_,z_];
Matrix T = output.multiply(input.transpose());

After this step, I multiply the coordinates of a point p that sit in plane P with matrix T. Of course, before that, I take the vector difference p - O.
The result should be the transformed point minus target origin (q - O_).

Hence, for each p in P, I transform p to its place in Q, which is denoted by q using the following computation:

q = T.multiply(p.subtract(O)).add(O_);

But what I get is, the reflection of the points through an imaginary (I don't know which) axis.

Am I missing some step? Is there a wrong computation?


I don't know if this is a typo in your question or the actual mistake:

x = normalize(Oa);
v = normalize(Ob);
z = a.crossProduct(b);
y = x.crossProduct(z);

For one you don't seem to be using v at all, but your z looks wrong to me. Did you mean the following?

z = normalize(Oa.crossProduct(Ob));

Note the Oa and Ob instead of a and b. The difference is the preserved perpendicularity of z on the plane P if P is moved away from the origin. I guess you wanted to do that through p.subtract(O) but that doesn't suffice.

Also note that I normalized z. This will result in orthonormal XYZ on both planes P and Q and means that scale is preserved. If the triangles abc and def are not similar (i.e. one is stretched) then T(P) = Q but T(a) != d. Don't normalize anything if you want to preserve stretching.

  • \$\begingroup\$ Yes, it was a typo. I corrected it. You suggest that I don't normalize the vector don't you? \$\endgroup\$ – padawan Jul 11 '14 at 23:15
  • \$\begingroup\$ It depends on what you want to achieve. T(P) = Q only, or that and T(abc) = def. \$\endgroup\$ – Jonas Bötel Jul 12 '14 at 1:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.