get track positions from trackerGNN

In this example:

https://www.mathworks.com/help/fusion/ug/passive-ranging-using-a-single-maneuvering-sensor.html

how do you extract the track coordinates and uncertainties calculated from the tracker?

I tried this:

[positions, covariances] = getTrackPositions(allTracks,'constvel')

but positions was a 3x1 array.

Answer 1

[Prashant Arora]

Hi,

For the Cartesian EKF, you should be able to use getTrackPositions(tracks, "constvel") and the output positions represents the position as 3xN array, where N is the number of tracks.

For the MSC-EKF, there's a helper function getTrackPositionsMSC defined in the example. You should be able to use that to get the position and covariance of the tracks.

Thanks,

Prashant

[black cat]

Thank you for the reply. After I ran the entire example and saved getTrackPositionsMSC into a separate .m file, I tried

[pos,cov] = getTrackPositionsMSC(allTracks(1),[0,0,0]');

But pos was still just a 3-element array.

[Prashant Arora]

Could you clarify what you meant by track coordinates and uncertainties?

The position output, pos, represents the position of the track as [x y z], and is expected to be a 3-element array. The covariance is a 3x3xN matrix representing the covariance of the error.

[black cat]

I would like to get the 3D positions of the track over time as estimated by the tracker over the course of the collection. The detections are angle-only.

[Prashant Arora]

The "tracks" output from the tracker represents the estimate of the object at the current time. It does not represent the entire history of the object estimate. To get the estimate over a course of time, you can use the getTrackPositions or getTrackPositionsMSC function inside the loop.

Each element of the "tracks" has a TrackID which represents the unique identity of the object. The code below shows an example of logging positions of tracks using a struct, however, you can use any format to store this information. This approach should work for multiple tracks, but can be simplied for single track use-case.

% Use a struct to store track-based position log. 
positionLog = repmat(struct('TrackID',0,...
    'Position',zeros(0,3),...
    'PositionCovariance',zeros(3,3,0)),0,1);

while advance(scenario)
    % Code to simulate scenario and generate detections
    
    
    % The output is the confirmed tracks at current time
    tracks = tracker(detections, time);
    
    % This represent the current positions and covariances of all the confirmed tracks
    [pos, cov] = getTrackPositions(tracks, 'constvel');
    for k = 1:numel(tracks)
        idx = find([positionLog.TrackID] == tracks(k).TrackID);
        if ~isempty(idx) % Track exists in the log
            positionLog(idx).Position = [positionLog(idx).Position;pos(k,:)];
            positionLog(idx).PositionCovariance = cat(3,positionLog(idx).PositionCovariance,cov(:,:,k));
        else % Add new track to the log
            newTrack.TrackID = tracks(k).TrackID;
            newTrack.Position = pos(k,:);
            newTrack.PositionCovariance = cov(:,:,k);
            positionLog = [positionLog;newTrack];
        end
    end
end

[black cat]

I see, thank you so much for explaining!