simtk.org

Hi,

I have a question related to Adjustment Model Markers.

I’m working with the arm26 model. I have collected experimental data for the static pose (it’s the default pose in the model).

Firstly, I’ve scaled the model using measurements from the experimental data.

Then, I want to adjust the markers. I’ve seen in another post, that you can tick the scale option and put manually scale factors all equal to 1.0, and tick the adjust markers option, also with the preview static pose, and do some trials until you get an error less than 1 or 2 cm (for the marker error, and the RMS).

If I’m not wrong, you can “help” the Scale Tool by placing the virtual markers closer to the experimental markers. I have done this directly in the GUI, picking up each marker one by one, and placing them near their experimental marker. You can also change the weight value for each marker. When I have changed the position of the virtual markers, the errors have diminished more than when I’ve changed the weight values.

I have three questions:

1. It’s the same (it has the same consequences) moving virtual markers and changing weight values for the markers, or it’s better one option than the other, or it’s ok to combine both options?

2. When I “untick” the preview option and I run the adjustment, the model that I get has the markers placed “between” the virtual markers that I have moved manually and the experimental values in the .trc file?

3. I thought that the weights were from 0 to 1, but I’ve seen in the example that there are also 1000, how can I know which markers will need greatest weights? Are the weight values only relative to the others? I mean, it’s the same to have 1 and 2, than 10 and 20?

Thank you,
Miriam Febrer

Hi Miriam,
Thanks for your detailed question, hopefully we can help.

The scale tool is really made up of two tools 1) Model scaler 2) Marker placer. The marker placer tool has a further two parts to it;

1)Marker Placer (first) runs an IK trial using your scaled model and your static pose data. IK adjusts model coordinates to satisfy a least-squares minimization of weighted marker errors. as the marker error between experimental and Model virtual markers determines the model coordinates, its important to highly weight markers you have 'confidence' in.These would typically be markers over bony landmarks (condyles) that can be easily registered between subject and model. The the marker weightings are represented as a ratio so a marker weighting of 100 versus 10 would be like saying 'Im ten times more confident in this marker position'. There are no hard and fast rules on doing this successfully, but some guidelines can be found on the OpenSim confluence website, here; http://simtk-confluence.stanford.edu:80 ... th+Scaling

I wouldnt get hung up on marker errors that you DONT have confidence in. If you misplaced a virtual marker on the model then you would expect (even hope) that the reported error would be large. What would be worrying is if a marker you have confidence in, and weighted highly, had a large error.

By manually adjusting markers prior to the IK step, to match experimental conditions, you have have a better chance of having valid results. You can also enter joint angles (that you have confidence in) and constrain the solution to match those angles. An example of this can be found here:
http://www.academia.edu/4169369/Automat ... noid_robot

2)The second step of Marker placer moves the virtual model markers the experimental marker position. When you untick a marker, it doesnt get moved during this step.

Hope that helps Miriam. Let us know how it goes,
-james

Thanks, James,
I've read the paper you have told me, and I'm working on the scale and adjusting tool... I'll tell you how it goes.

Miriam

Hey Miriam,

I am a new user of OpenSim. I am kind of confused about some words. The virtual markers, are they those markers on the model (for example, on the sample model Gait2354), and the experimental markers are those what we used during the measurement with Motion tracking system, am I right? I also went through the paper mentioned above, but I have a question, how to perform the auto-registration? Thank you.

miriamfebrer wrote:Thanks, James,
I've read the paper you have told me, and I'm working on the scale and adjusting tool... I'll tell you how it goes.

Miriam