In target tracking and sensor fusion, is it possible to keep state vector in cartesian coordinates while keeping measurements in spherical coordinates?

I want to track a target which is onserved by a radar that gives measurement in spherical coordinates(range, azimuth and elevation). However, i want to keep, predict and correct states in cartesian coordinates. For example, is that possible create a H matrix which is multiplied state in cartesian coordinate and give measurement prediction according to that while using extended kalman filter? If it is possible, can I find an example related to this issue in sensor fusion and tracking toolbox? Thanks in advance.

Answer 1

[Elad Kivelevitch]

Yes, it is possible.

Based on your previous question, I am assuming you're using one of the available trackers (e.g., trackerGNN) with a FilterInitializationFcn that returns a trackingEKF object, e.g., initcvekf.

To use the tracker, you must format your radar detections into an objectDetection format. Use this example (especially, the third section) to see how to define detections in a spherical frame.

https://www.mathworks.com/help/fusion/ug/convert-detections-into-objectDetection-format.html

Our built-in models, for example the constant velocity model, all support measurements in spherical coordinates and state definition in rectangular frame, which is exactly what you're looking for.

[Aatif Sulaimani]

Hi Elad,

I am also trying to do the same but I do not want to use the conversion through objectDetection. I want to convert measurement using measurement function in trackingEKF object. Is it possible? Is there any reference on how I can implement it?

Thank you!