Coordinates of object relative to other object

[ohmboy]

Aloha everybody!

i am having major difficulties concerning this thing and would dearly appreciate if someone would be willing to help me out...

ok, here goes.
I need to get the coordinates of object1 in a second object's (lets call it object2) local coordinate space. Initially i thought that by simply parenting object1 under object2 i would get the correct values from object1's translate values. However, this is only true if object2 has not undergone scaling nor rotation.
What do i need to do to take scaling and rotation into account?
Does there exist a MEL command that gives me the coordinates of one object with respect to another objects local coordinate frame?
I basically need to get the coordinates of object1's pivot point with respect to object2's local axis...
Or will i need to write a script that does a coordinate transformation? How can i then get the object's local axis?

any help much appreciated... Thanks!!!

alexander

[joojaa]

pointmatrix multiply 0 0 0 (or the actual point in question) with the inverse worldmatrix matrixmultiplied with target objects world matrix

Or point constrain if you need to querry the data more than once interactively.

[ohmboy]

Thanks Joojaa for your reply, but alas i cant get it working properly...
I'll explain what i do precisly...

First of all, i am using Bryan Ewert's MATRIXPLAYGROUND script for all the matrix stuff (which i found on this site), so i dont assume anything is wrong with taking the inverse and stuff like that...

To conceptualize it a bit more, consider pCube1 placed at {5,5,5} without any rotation nor scaling applied and pCube2 placed at {-5,-5,-5}. This means that the result, i.e. the coordinates of pCube2 with respects to pCube1's local coordinate frame should be {-10,-10,-10}

I verify this as follows in MEL:
//get worldMatrix of pCube1
float $worldArrayCube1[] = xform -q -m pCube1;
//convert this float array to a matrix using Ewerts tool
matrix $worldMatrixCube1[4][4] = doubleToMatrix44($worldArrayCube1);
//take the inverse of this matrix
matrix $inverseOfWMCube1[4][4] = inverse($worldMatrixCube1,4);

//get the world matrix of pCube2
float $worldArrayCube2[] = xform -q -m pCube2;
matrix $worldMatrixCube2[4][4] = doubleToMatrix44($worldArrayCube2);

//now multiply both matrices, i.e. multiply the inverse worldMatrix of pCube1 with the worldMatrix of pCube2

matrix $multiplied[4][4] = multiplyMatrix($inverseOfWMCube1, $worldMatrixCube2);

Now, in the last row of this matrix should be the coordinates of pCube2 with respect to the local coordinate system of pCube1.
All is fine as long as i dont rotate or scale pCube1. The moment i do scale or rotate pCube1, its all wrong.
Say for instance that i keep the location of both cubes at respectively {5,5,5} and {-5,-5,-5} and i rotate pCube1 over 90degrees around the Y axis. The correct result should be {+10,-10,-10} but instead i get {0,-10-10}...

What is going on here?
Oh, and pointMatrixMult with {0,0,0} or pCube2's translate coordinates does not amount to anything usefull...

thanks for your help!

hope someone can point me out in the correct direction (again).

A

[campbell]

how about point mult with the inverse matrix? do a serries of these trasformations if the target space is stacked.

[ohmboy]

Hi Campell,

could you give me an example of what you mean, cause i dont really understand what you are talking about...
i.e. point multiple which point with which inverse matrix?
Do i need to pointMatrixMult the translate coordinates of the object whose coordinates i want with the inverse worldMatrix of the object whose local coordinate space i use. When i do this i get wrong values though...

This thing sounds simple enough but for some reason i just cant manage to get the correct values when the objects have undergone scaling and rotation...

AAAaaarrrggghhh!!

please help... i would be forever in your debt.

alexander

[joojaa]

use node networks its faster since pointmatrix mult allready uses them... and its defined a bit wonkily (its not a point matrix multiplication at all infact read the mel procedure, so no wender it yelds the worng result) so its better to use constraints since the computation is WAAAAY much faster.

[joojaa]

heres a example (for a empty scene jsut because im lazy not wanting to encapsulate it all):

polyCube -w 1 -h 1 -d 1 -sx 1 -sy 1 -sz 1 -ax 0 1 0 -tx 1 -ch 1;
polyCube -w 1 -h 1 -d 1 -sx 1 -sy 1 -sz 1 -ax 0 1 0 -tx 1 -ch 1;
setAttr "pCube1.translateX" 5;
setAttr "pCube1.translateY" 5;
setAttr "pCube1.translateZ" 5;
setAttr "pCube2.translateX" -5;
setAttr "pCube2.translateY" -5;
setAttr "pCube2.translateZ" -5;
createNode multMatrix;
connectAttr -f pCube1.inverseMatrix multMatrix1.matrixIn[1];
connectAttr -f pCube2.matrix multMatrix1.matrixIn[0];
createNode pointMatrixMult;
connectAttr -f multMatrix1.matrixSum pointMatrixMult1.inMatrix;
getAttr pointMatrixMult1.output;

just querry the last point everytime you need the answer using getAttr pointMatrixMult1.output; altough theres no reason you cant connect it to some ui or atributes for covinient reading.

alos this is MUCH MUCH MUCH faster than yusing matrixplayground

altough point constrain aslo yields a slightly faster computation. So use a parented pointconstrain instead if you realy need this data.

By the way thanx for the tip that the mel command is a actual porcedure wich instead uses a pointMatrixMult node wich is in vector multiplication mode wich is not what one realy expects. so better build nodenetworks instead.

Anyway i never actualy use this since i can allways use nodenetworks instead (because of the performance boost.) Also because creating and deleting nodes is a bit insane in my oppinion one could allways recycle them tough.

blooper:
oops there was a error in the connection forgot that matrix mult is not comutational so order does count correction in code in blue.

[campbell]

i ran into that pointMatrixMult the other day too whilst looking at that script, and had exactly the same thoughts.

hey joojaa is there any benefit of your method over just doing a point matrix mult on the target's inverse matrix?

i like the constraint method, its got that nice agricultural feel to it, the way i like it.

[joojaa]

Im not sure what you mean exactly but just using inverse and point matrix mult does work if they are parented to eachother offcourse but this is a general coversion form any space to any space.

not realy constraint is slightly faster but the point is offcourse it can be done.

But this yelds a faster orient and point constrain combination. Certain thigs are easier to Tweak with asetup like ths tough.

[campbell]

mmm, ok - so they validly acheive the same result, just by different means? to deal with differing parent spaces, they both require serrialising. i guess one adavantage of doing this by matrix multing, in a node context, is that you can do stacked spaces in one node, rather than a string of point matrix mults.