Background
Model View Transformation
OpenGL does not have a view matrix, but only a model view matrix. Generally, game engines abstract the view matrix.
The overall transformation is:
V' = projection * view * mode * V;
Parent-Child node Transformation
V' = worldmatrix * V = (parentworldmatrix * localmatrix) * V;
Skeleton Transformation
In fact, the joints actually represent, rather than the bones, but the bones are better understood.
By default, the coordinates of the pose Grid Object are converted to the local coordinate space (offsetmatrix) of the skeleton ).
After that, it is transformed from the skeleton space to the animation posture world coordinate (nodewroldmatrix ).
V' = nodewroldmatrix * offsetmatrix * V;
Here there is a bindshapematrix which does not know what it means. It is estimated that it is the offset when the mesh object is bound to the skeleton, that is:
Offsetmatrix = inversewroldmatrix (inverse world transformation) * bindshapematrix;
Because the MVP (mode view matrix) matrix already contains the M matrix, you must exclude the top-level transformation in the nodewroldmatrix or change the MVP to VP.
V '= projection * view * nodewroldmatrix * offsetmatrix * V;
Skin Transform
Skin transformation, the final matrix of each nodewroldmatrix * offsetmatrix is called a matrix palette.
Each skin vertex is associated with multiple bones, that is, the matrix in multiple matrix palette. Generally, four integer indexes are used for this purpose.
Multiply the matrix of each matrix palette, and then take the weighted average value, that is, the final position.
Vomit time
I have guessed most of these concepts, and I cannot find any such information.