Page 1 of 1

Tracking Data CSV

Posted: Tue Aug 07, 2012 7:30 am
by Asa_E-E
I am working to replicate the CSV for Tracking data that is exported from Tracking Tools and have some questions about some of the data.

Righthanded/Lefthanded: Is there a function that can be used to check this or is this something that will need to be manually set depending on how my coordinate systems is defined? I imagine this is set in the calibration file but I don't see anyway of interacting with a calibration file besides opening it.

Marker Tracked: I see a function for checking if a trackable is tracked, but nothing for a marker. Is the assumption that if the trackable is tracked, all the markers in that trackable are tracked?

Marker quality: How is this calculated?

Mean Error: How is this calculated?

Marker ID: Frame Line - I'm using TT_FrameMarkerLabel() to get the Marker ID. Trackable Line - Use the same function?

Time Stamp - When does TT_FrameTimeStamp() begin keeping track of time? At the first TT_Update()?

Thanks,
Asa

Re: Tracking Data CSV

Posted: Tue Aug 07, 2012 11:50 am
by NaturalPoint - Mike
Hello -

To answer your questions:

Handedness can be set in the GUI by going to edit - tracking tools settings - Options Tab. The option is listed as "Right handed coordinate system" and can be set to either true or false to switch between left and right handed systems. It is stored in the project file, as are all settings. Calibration files only include camera intrinsic and extrinsic information as well as the origin and global plane orientation.

Tracked: Rigid bodies will give you a tracked indicator that tells you if it's currently found in the point cloud. Depending on your rigid body composition and settings, you may or may not need all the points to track the rigid body. The Rigid body line will include the projected marker locations based on the solved centroid position and orientation, as well as the point cloud markers used to solve. If there are less than the expected amount of point cloud markers, you can find the missing one by the difference between the two sets.

Marker quality: The method of this measurement is proprietary, however it is a good metric to use, it's measured between 0 and 1.

Mean error is calculated as the potential difference between calculated centroid and potential location based on the positions of the camera 2d data used to calculate it. It's presented in mm.

Marker ID frame line/trackable line: I'm unsure of what you mean, the trackable line contains both the point cloud marker location and the expected location derived from the centroid location/orientation. You would use it for the actual point cloud marker, but for the expected position (or best fit position) you would get the centroid of the trackable followed by the marker offset. See: TT_TrackableLocation(); TT_TrackableMarker(); TT_TrackablePointCloudMarker();

The time begins with TT_Initialize();

Re: Tracking Data CSV

Posted: Tue Aug 07, 2012 1:43 pm
by Asa_E-E
Marker Quality: Is there a function that allows me to access the marker quality through the API?

Marker ID: In the CSV from Tracking tools, there are two points where it says Marker ID. In the Frame line, it is Marker Detail (x,y,z,id). I'm using TT_FrameMarkerLabel() to get this ID. In the Trackable line, it is Trackable Marker Location and ID (for each marker). I don't see any other functions for the marker ID besides the FrameMarkerLabel one, but I'm wondering if that returns the proper ID for the Trackable line. Does that clarify the question?

Re: Tracking Data CSV

Posted: Tue Aug 07, 2012 3:12 pm
by NaturalPoint - Mike
The Trackable Marker Location in the trackable line isn't the one with the marker ID. The ID is on the Point cloud marker location. Here's the difference:

The trackable is solved to a best fit, and often isn't exactly on top of the point cloud markers to a margin of error of 0. Because of this, and because we allow a certain degree of flexibility in the solving parameters, we first give you the trackable marker location, which is where the markers would theoretically be in a perfect world if the point cloud markers were exactly where they should be. After listing those, we provide the actual point cloud marker locations and the ID (which can be compared to the complete list of markers in the frame line). It's literally a repetition of that data, but with the markers that have nothing to do with that rigid body stripped out of it.

There is no way to get marker quality out of the API. It's a metric that's applied to the data during export as part of the export process.