Which tracker from the sensor fusion and tracking toolbox suits best for tracking cars?

Hello,

since some time I am trying to track cars where I measure the position and the speed every second and feed that information to a tracker. At the end I want to have all the tracks of the cars so that I can compare it to the real ones. So far I used the trackerTOMHT and trackerJPDA but without any success and I am kinca stuck.

My question is divided in two parts:

1. How would you approach such a tracking domain? Would you use different trackers?
2. Do you have any hints or personal experience how to improve the settings of one of the trackers I used so far?

Thank you!

Answer 1

[Alan]

Hi,

I understand that you want to track the position and velocity of vehicles from a video source.

1. It looks like you are using a multiple object tracker with multiple sources. You could check out the following example which uses a single source: https://www.mathworks.com/help/driving/ug/track-multiple-vehicles-using-a-camera.html#TrackMultipleVehiclesUsingCameraExample-2
2. The velocity of each vehicle can be obtained from their tracks using getTrackVelocities: https://www.mathworks.com/help/driving/ref/gettrackvelocities.html#mw_0f6140c3-92aa-4a84-991e-fe45a431b770

Coming to your question of using different trackers, I’m not sure whether you mean by using different trackers for different objects or different trackers for the same object.

1. In case of the former, you might have to find a way to dynamically determine which tracker is better based on the properties of each detected object.
2. In case of the latter, you might have to figure out a way to combine the estimation, assignment, updating, and deletion functionalities of all the trackers assigned to the detected object, which might get complicated to implement.