Let C be a right-handed three-dimensional coordinate system.
Let R be a linear transformation that preserves an orthonormal basis, such as a rotation around or reflection through the origin.
Let R be applied to any point in C, say P, and let Q be the image of P.
R : P ↦ Q
Let R be applied to the orthonormal basis vectors that define the axis directions of C to give a new coordinate system C'.
R : C ↦ C'
In the diagrams below R is shown as a π/2 radians rotation about the y-axis:
-
Coordinate system C
y-axis | / Q | / | | / | | / | | / | | / / | / / |/ / _/_ _ _ _ _ _ _/_ _ _ _ _ _ O/ x-axis / P / | / | / | /- - - - - - -| / / z-axis
-
Coordinate system C'
y'-axis x'-axis | / | / | / Q' | / | | / | | / | | / | | / / | / / |/ / _/_ _ _ _ _ _ _/_ _ _ _ _ _ O/ z'-axis / P' / | / | / | /- - - - - - -| / /
In the diagrams:
- O is the origin and does not move.
- R maps the x/y/z axes to the x'/y'/z' axes respectively.
- R⁻¹ (the inverse of R) transforms Q to P in C and therefore transforms Q' to P' in C'.
Suppose we wish to find the coordinates of P in C', that is we wish to find P'. (This is the change of basis we require.)
We have that
R⁻¹: Q' ↦ P'
Since Q' has the same coordinates in C' as P in C, to find P' we simply need to apply R⁻¹ to P:
Q' == P ∴ R⁻¹: P ↦ P'
The columns of the matrix R are given by the basis vectors of C' in C:
let R = |a d g| then |1| |1| |a|, |0| |0| |d|, |0| |0| |g|
|b e h| |0|→ R|0| = |b| |1|→ R|1| = |e| |0|→ R|0| = |h|
|c f i| |0| |0| |c| |0| |0| |f| |1| |1| |i|
A matrix whose columns constitute an orthonormal basis is an orthogonal matrix. The inverse of an orthogonal matrix is its transpose. That is
R⁻¹ = Rᵀ
and so
Rᵀ: P ↦ P'