Experimental and model markers are too far apart after scaling

[Cai Birch]

I've been trying to use OpenSim for some months now and I'm struggling to get my model to scale satisfactorly. I've read and watched all the tutorial/webinar sessions but they offer very limited advice for troubleshooting issues. Several markers do not perfectly overlap once scaling is complete. In fact all model markers that are 'fixed' to bony landmarks e.g. RASIS, Knee Epicondyles, Acromions, etc. never perfectly match the experimental marker locations. I have exhausted all of my ideas on how to remedy the problem...Modelling with a poorly scaled model will yield wildly inaccurate results, so it seems like an important area to get right. Could my input data be a culprit? The model is always rotated 90 degrees when scaled, could the different coordinate systems used by Vicon and Opensim be to blame?

Many thanks

Scaling failed.jpg (91.57 KiB) Viewed 346 times

[jimmy d]

Several markers do not perfectly overlap once scaling is complete.

Are the model markers set to fixed? If so they won't be moved to match the experimental position.

In fact all model markers that are 'fixed' to bony landmarks e.g. RASIS, Knee Epicondyles, Acromions, etc. never perfectly match the experimental marker locations.

Why would you expect them to 'perfectly' match the experimental locations? There is error in the location of the experimental marker on your subject, error in the geometry of the model, error in your scaling, error in you model marker placement. The only way to perfectly match the markers is to do an MRI/fluro scan of the subject while wearing the markers. That way you would be able to produce subject-specific models with precisely registered markers.

Modelling with a poorly scaled model will yield wildly inaccurate results, so it seems like an important area to get right.

There are numerous sources of error in motion modeling, scaling is just one of them. I suggest studying up on the sources of motion capture error and figure out how you can answer your question given the sources of error. A good place to start is the four-part series 'Human movement analysis using stereophotogrammetry' in Gait and Posture.

The model is always rotated 90 degrees when scaled, could the different coordinate systems used by Vicon and Opensim be to blame?

IK is used to pose the model during the marker registration step. This moves the model to the location of your experimental data. So either you are not rotating your data correctly, or you are rotating correctly and your subject was 90 degrees rotated when you took the static trial.