Autodesk Forge模型变换矩阵：placementTransform与refPointTransform用途咨询
Hey there! Great question—these two transforms are core to 3D geometry pipelines, so it makes total sense you’d be curious about their extra parameters beyond basic translation/rotation. Let’s break them down clearly:
This is a flattened 3x4 affine transformation matrix—since 3 rows × 4 columns = 12 elements, it’s stored as a 1D vector for efficiency. Here’s how the 12 parameters map:
- The first 9 values make up a 3x3 rotation matrix (handles Rx, Ry, Rz rotations; it’s usually orthonormal for pure rotation, but can also include shearing if needed)
- The last 3 values are the translation vector (Tx, Ty, Tz)
What it’s used for:
- Defines an object’s absolute position and orientation in a 3D world space. It transforms points from the object’s local coordinate system to the global/world coordinate system.
- Super common in 3D rendering, AR/VR (placing virtual objects in the real world), and CAD tools where you need to position models precisely without worrying about scaling or other complex transforms.
This is a flattened 4x4 homogeneous transformation matrix—4 rows × 4 columns = 16 elements. Homogeneous matrices let you represent all affine transforms (rotation, translation, scaling) in a single structure, plus handle perspective if needed. The breakdown is:
- The top-left 3x3 submatrix: handles rotation, scaling, and shearing
- The first three values of the 4th column: translation (Tx, Ty, Tz)
- The 4th row is almost always
[0, 0, 0, 1]for standard affine transforms (this ensures points stay in 3D space when transformed)
What it’s used for:
- Manages transforms related to an object’s reference points (like its center of mass, anchor point, or a specific feature you want to align). For example, if you need to shift an object so its base reference point matches a target location, this matrix encodes that full transform.
- Used in hierarchical 3D scenes (like parent-child object relationships) to combine transforms from parent objects down to children. It’s also essential when you need to include scaling in your transform—something the 3x4 placementTransform can’t handle natively.
To circle back to your initial thought: you’re totally right that both tie into translation/rotation! The "extra" parameters are just the full components of rotation matrices (a 3x3 matrix has 9 elements, not just 3 rotation angles) and the homogeneous matrix’s extra rows/columns for scaling and coordinate system consistency.
内容的提问来源于stack exchange，提问作者Fred Coud
