Trackmate - Tracking at High Speed




I’m analyzing images acquired that are about flow in pipes containing markers, and I’m trying to track the markers using the Trackmate plugin.
The Trackmate is actually doing great when the flow speed is low, however, when the speed is too high the LAP tracker links the marker to the neighboring marker because it is closer than the same marker spot in the following frame!.

I’ve tried to modify the tracks manually using the TrackScheme, but with more than 5000 frames in each data set, it is a very difficult mission and I assume it is not the optimal way to do it.

I also think that assigning a specific trajectory for the tracking (along the y-axis of the pipe for example) or adding an angle parameter to the linking will be useful, instead of just depending on the distance to construct a link, but my coding knowledge is limited.

(I’m using The LoG detector and The LAP Tracker in the setup).
Attached you can find an image of a tracking done for just 200 frames, the color map is for Mean velocity, as you can see the markers in the middle of the pip[e are not well tracked due to high speed, however, the markers on the side of the pipe are better tracked due to a lower speed.

I need advise on this problem, please.

Thank you very much for helping.


Hello Abdelrahman,

Your images typically fall into a category of problems in which TrackMate LAP tracker won’t be working optimally. You need to input a large search radius to accommodate long jumps from frame to frame. But then, linking to another, closer particle becomes possible, because the LAP tracker does not include a priori information such as constrained flow.

The dummy advice I can give is to increase the imaging rate so high that even the quickest particle only move by a little from one frame to the next one. But I guess that with high speed flow it might not be doable.

Otherwise, I would recommend you to try to Kalman tracker of TrackMate:

This tracker tries to link particles assuming they have a roughly constant velocity vector, which is just your case.

The key difficulty is with track initialization: if the tracker can link correctly the first two particles of a track, it will probably manage to link all the other ones, correctly extrapolating their position with the measured velocity.But the track initiation is crucial, so I would recommend playing with the Initial search radius settings to get it right.

Otherwise if it does not work we would have to create a new tracker for your problem, but I get this is out of the picture for now.


Thank you very much for the quick reply.

I’ve been applying your recommendation since yesterday.
Increasing the imaging rate is not dummy advice at all, but unfortunately is not possible because of the camera readout.
The Kalman tracker (linear motion LAP tracker) works perfectly with the high speed, however, it is not working with the low-speed flow (maybe due to flow velocity change), but it is fine as I can use the LAP tracker for the low speed and get good tracking. :+1:

I really appreciate your kind and effective reply, Thank you!!


I am really glad it works!
Can you post a screenshot of the Kalman tracker results? It will be super instructive for others.


Sure. With pleasure!


I could totally print it on a metal poster and put it in the lab. Bravo!


I would like a metal copy of that as well!