Virtual Pelvis Markers are not Matching the Experimental Markers

[Adam Galloy]

Hello,
I tried scaling a full body model using marker data of someone standing still. The results I got had the pelvis of the model twisted around almost 90 degrees. I checked the models pelvis markers and compared them to the experimental markers using the Preview Experimental Data feature and they both seemed to agree. I tried doing a preview static pose and all of the markers except the pelvis markers (RASI, LASI, RPSI, and LPSI) were at least in a recognizable position. I tried scaling with the pelvis markers having a very high weight (to see if I could put them in the correct position) and still got the twisted hips. I tried another static pose preview and found that the virtual markers were being oriented incorrectly even though the weight was set high to match them. The rest of the markers are still at least look normal even if the positions of the markers are a little off. I have attached a picture of the last preview I talked about. The experimental markers are in blue and the virtual ones are pink. I only showed the RASI, LASI, RPSI, and LPSI markers for clarity.

Thanks,
Adam Galloy

[jimmy d]

To pose the model, a single frame IK solution is used. The behavior you are seeing is typically caused by incorrect labeling of markers, either experimentally or on the model.

[Adam Galloy]

I have checked marker labels both in model and in data and they match.

[jimmy d]

Please attach a .zip that includes the model (.osim) and marker files (.trc).

[Adam Galloy]

This folder has the model with markers attached, the collected marker data (DylanStatic), my Scale settings, and the original marker set (it was originally for a different model I have used that has a similar problem so I had to manually adjust/add many of the markers to the model maybe I did something wrong during that process).

Thanks,
Adam Galloy

[Deepak Jindal]

Hey Adam,
As i tried your problem it seems that you didn't transformed the data correctly from your laboratory axis to OPENSIM
axis correctly so please check proper transformation.

[Adam Galloy]

Really? I had thought about that but the only part of the body that isn't going to the right position is the pelvis and the experimental marker preview coordinates are all in the correct position when I load them. Do they have to perfectly overlap the initial model or can they be translated as they currently are? When I run Inverse Kinematics on the model I can get the model to walk and jump rather convincingly (except for the twisted pelvis) so I thought the transformation was correct. If you can provide any more explanation of what is not translated correctly that would be great!
Also, I know that the force plate data (if I even included it in the folder) is not transformed correctly. I haven't gotten around to that yet.
Thanks,
Adam Galloy

[Adam Galloy]

I collected data again with the same subject. This time I experimented by having him stand facing the positive direction and the negative direction. I found that if I have him face the positive direction the pelvis will not rotate but if he faces the negative direction the pelvis will rotate. I didn't experiment with having the subject stand in different locations (he was always standing at the origin of the lab's coordinate system). I figured that OpenSIm used the relative positions of the markers for analysis such as scaling and inverse kinematics so that it would not matter if the subject was facing a different direction or standing in a different location, but I appear to be wrong.

[Josh Baxter]

many of the default opensim models have ranges of 90 degrees set for the pelvis orientation coordinates. If you go into the osim file and change these ranges to <range>-3.14159 3.14159</range> the problem of lining up the subject with the lab coordinate system should be resolved.

[Adam Galloy]

Thanks for the reply Josh!
I meant to send a post here mentioning that I had found that out while running Inverse Kinematics on a person turning around. I have since fixed this in my models so that they are able to move more freely.

Adam