Hi there,
I am a new user of OpenSim. I read that OpenSim needs at least three markers to track the 6dof motion of a body segment. If this is not the case (i.e marker set of Vicon PlugIn Gait does not fulfill this requirement), the results from OpenSim IK and ID can be valid or are by definition incorrect?
Thank you,
Nikos
Can the Vicon PluginGait marker set be used in OpenSim?
-
Nikolaos Karavas
- Posts: 2
- Joined: Mon May 27, 2013 10:08 am
-
David John Saxby
- Posts: 83
- Joined: Mon May 09, 2011 8:39 pm
Re: Can the Vicon PluginGait marker set be used in OpenSim?
Hi,
I believe you will need at least 3 markers per segment/body to do ANY 3D motion analysis. This is not a unique demand of OpenSim.
Plug-in gait has 3 markers on each segment. Otherwise it cannot compute direct kinematics or ID.
David
I believe you will need at least 3 markers per segment/body to do ANY 3D motion analysis. This is not a unique demand of OpenSim.
Plug-in gait has 3 markers on each segment. Otherwise it cannot compute direct kinematics or ID.
David
-
Nikolaos Karavas
- Posts: 2
- Joined: Mon May 27, 2013 10:08 am
Re: Can the Vicon PluginGait marker set be used in OpenSim?
Hi David,
thank you a lot for your reply.
I agree that the PIG needs 3 markers per segment to compute the kinematics. However, as it can be seen in the corresponding vicon skeleton template (.vst file) these markers are not necessary real. PIG modeling takes the real marker trajectories and generates virtual marker trajectories based on some assumptions and specific anthropometric measurements. For instance, it omits the knee medial marker but it needs the knee width measurement.
Hence, my question is if OpenSim does something similar.
best,
Nikos
thank you a lot for your reply.
I agree that the PIG needs 3 markers per segment to compute the kinematics. However, as it can be seen in the corresponding vicon skeleton template (.vst file) these markers are not necessary real. PIG modeling takes the real marker trajectories and generates virtual marker trajectories based on some assumptions and specific anthropometric measurements. For instance, it omits the knee medial marker but it needs the knee width measurement.
Hence, my question is if OpenSim does something similar.
best,
Nikos
-
Ton van den Bogert
- Posts: 167
- Joined: Thu Apr 27, 2006 11:37 am
Re: Can the Vicon PluginGait marker set be used in OpenSim?
Opensim uses full-body inverse kinematics (IK) and does not need three markers on each segment. It estimates all kinematic variables simultaneously by minimizing the sum of squared distances between simulated marker positions and measured marker positions.
Most joints do not have 6 DOF so a surprisingly small markerset can be sufficient for this to work. It is not possible to say exactly how many markers are needed. Typically, the markerset is designed to be redundant, and the kinematic analysis can still work even if a number of markers drop out.
This avoids the need for "virtual markers" which is not elegant and can introduce error. A disadvantage of IK is that only the kinematic degrees of freedom of the model are estimated. If the model has 1 DOF (flexion-extension) at the knee, that is the only knee motion variable you will get out of the kinematic analysis. On the other hand, this limitation is exactly what makes the results suitable for driving a musculoskeletal model or a graphics model. The joints won't come apart as often happens in a 6-DOF analysis with 3 markers per segment.
The Opensim documentation will probably have some references to the algorithm.
Some history for those who are interested. I implemented this algorithm in 1996-1997 for Motion Analysis Corp and the software was for sale in 1997 as "Mocap Solver" for the animation industry. Others were working on this idea at the same time. Lu & O'Connor at Oxford were the first to publish it (http://www.ncbi.nlm.nih.gov/pubmed/10052917) and Zoran Popovic at Carnegie Mellon University described it in one paragraph in his thesis (http://reports-archive.adm.cs.cmu.edu/a ... 99-106.pdf, page 77). We all discovered it independently, it seems. There may be earlier uses that I don't know about.
Motion Analysis included the algorithm in the SIMM software and then Opensim had it from the start because the same capability was needed there. You can find the least squares IK algorithm now also in Anybody and Visual3D and in Human Body Model from Motek Medical (disclosure: I was a developer of that).
So, three markers per segment is no longer the only way to do kinematic analysis. If you don't have 3 markers per segment, a model-based IK will be far better than creating virtual markers. If you do that, you have no idea how much of each kinematic variable is really measured and how much is assumed. With a model, on the other hand, you are perfectly aware of the assumptions of how each joint can move.
The PiG markerset should be more than sufficient to do kinematic analysis with an Opensim full body or lower extremity model. If I remember correctly, there are tutorials for how to attached a markerset to a model.
Ton van den Bogert
Most joints do not have 6 DOF so a surprisingly small markerset can be sufficient for this to work. It is not possible to say exactly how many markers are needed. Typically, the markerset is designed to be redundant, and the kinematic analysis can still work even if a number of markers drop out.
This avoids the need for "virtual markers" which is not elegant and can introduce error. A disadvantage of IK is that only the kinematic degrees of freedom of the model are estimated. If the model has 1 DOF (flexion-extension) at the knee, that is the only knee motion variable you will get out of the kinematic analysis. On the other hand, this limitation is exactly what makes the results suitable for driving a musculoskeletal model or a graphics model. The joints won't come apart as often happens in a 6-DOF analysis with 3 markers per segment.
The Opensim documentation will probably have some references to the algorithm.
Some history for those who are interested. I implemented this algorithm in 1996-1997 for Motion Analysis Corp and the software was for sale in 1997 as "Mocap Solver" for the animation industry. Others were working on this idea at the same time. Lu & O'Connor at Oxford were the first to publish it (http://www.ncbi.nlm.nih.gov/pubmed/10052917) and Zoran Popovic at Carnegie Mellon University described it in one paragraph in his thesis (http://reports-archive.adm.cs.cmu.edu/a ... 99-106.pdf, page 77). We all discovered it independently, it seems. There may be earlier uses that I don't know about.
Motion Analysis included the algorithm in the SIMM software and then Opensim had it from the start because the same capability was needed there. You can find the least squares IK algorithm now also in Anybody and Visual3D and in Human Body Model from Motek Medical (disclosure: I was a developer of that).
So, three markers per segment is no longer the only way to do kinematic analysis. If you don't have 3 markers per segment, a model-based IK will be far better than creating virtual markers. If you do that, you have no idea how much of each kinematic variable is really measured and how much is assumed. With a model, on the other hand, you are perfectly aware of the assumptions of how each joint can move.
The PiG markerset should be more than sufficient to do kinematic analysis with an Opensim full body or lower extremity model. If I remember correctly, there are tutorials for how to attached a markerset to a model.
Ton van den Bogert