Marker adjustment in IK part of Scaling

[Pouya Amiri]

Hi all,

I am new to open sim and have been trying to figure out how to work with it. I read the content of http://simtk-confluence.stanford.edu:80 ... th+Scaling and also watched the videos. My question is that after previewing the results of IK in scaling, how can we evaluate our results? Do we have to compare the posture of the model (joint angles) with, for example, the pictures of the real subject that we have? or we have to evaluate our results based on marker positions? What I think, is that we have to try to match the posture by changing virtual markers on the model (I mean giving priority to reproducing the posture, and then marker positoins). The problem is that the model is very sensitive to the position of the virtual markers, and by a small change in the position of virtual markers, the angles change a lot. It is also seen even when changing the positions of tracking markers on the segments. I tried to start from foot (with minimal markers placed on anatomical landmarkers) and go up to other segments and each step adjust the markers in order to get a reasonable posture for the contained segments. However, this method is also annoying, because when adding new markers, all the angles adjusted before, change again. and also if you start with the whole set of markers, the process is the same and complex.
This is a big problem especially when you want to apply a model to a large number of subjects, meaning that each time you have to spend a lot of time to adjust the markers for each subject. for the anatomical landmark, it is easier because of the fidelity we have. however, for the tracking markers, usually there is no specific position constraint and it makes the problem harder.

I would appreciate any help and suggestion helping me to make this process easier, more reliable, and accurate.

Thanks in advance.

Pouya

[jimmy d]

hi-
The link is to an abstract at ISB that dealt with this issue. Hope it helps.
http://www.academia.edu/4169369/Automat ... noid_robot
cheers,
-james