How can I use a modified measurement model or observation matrix H in trackingEKF?

I am tracking camera detections using trackerGNN, trackerJPDA and trackerTOMHT and with the help of trackingEKF object. I would like to have my observation matrix H modified to my needs. To begin, I have camera detection as x y pixel location. Now I would like to project it on a ground plane image using homography matrix. I have the homography matrix and I want to use it in observation matrix (H) of Kalman filter to map my measurement to the state which is in ground plane as a measurement function. My state is [x vx y vy] where x and y are pixel location in ground plane and measurement [x y] where x and y are pixel location in camera image. As my homography matrix is [h11 h12 h13; h21 h22 h23; h31 h32 h33] I cannot just use it directly as the equations would be h11x + h12y + h13 , h21x + h22y + h23 and h31x + h32y + h33. After that I divide 1 and 2 by 3 value to normalize it. Can anyone let me know how this is possible?

[Walter Roberson]

@Aatif Sulaimani

You flagged your own question as Unclear. As the person who posted the question, you are the person with the greatest chance of correctly editting the question to make it clearer.

Answer 1

[Elad Kivelevitch]

Sounds like you should use a trackingEKF and create a measurement function that converts your (x,y) pixel measurements.