Feature Request: Combined LAP Kalman tracking #315
Description
I am noticing that the Kalman tracker and LAP trackers both have advantages and disadvantages. The kalman gives me nice long continuous traces, but they sometimes contain false positive sections where they have included noise as part of the trace. The LAP picks up traces but they are often disconnected. I have played with the parameter for both and and reasonably sure they are correct. It would be fantastic if the results of both trackers where "combined".
Would it be possible to run them both and then add a filtering option so based on track proximity and track tortuosity (how curvy the line is)? Sorry if that's worded poorly I have outlined the process in my head below in what I hope makes more sense. I know its much easier said than done so if this even gets considered I would be very happy!
- Load data and run spot detection
- Run LAP tracking and filter traces
- Run Kalman tracker and filter traces
- Combine resulting tracks
- Confirm which tracks are duplicates (i.e the same track has been detected in both tracking approaches) based on if the same spots have been chosen during the tracking process, track proximity, overlap, tortuosity.
- Filter duplicate tracks based on which has the best track mean quality
I would do it my self but I have no idea how to program anything. I know this is a big ask and its much easier said than done so if this even gets considered I would be very happy!