It's quite understandable to do pose estimation of Fiducial Markers and also track with quite a good accuracy. So, I have relatively broad questions:
- Sometimes when any occlusion occurs, even with a finger, the tracker stops detecting.
- I have the coordinates and the center point of the aruco marker. It's quite easy to find that. But using those coordinates, how do I predict the values ahead of that center coordinate?
- And while I am able to detect the predicted point and generating the bounding box to track the aruco with such prediction, how do I predict the speed as well?
I have been searching the whole internet to understand how Kalman Filters, UKF(which is the actual requirement) and EKF actually work. The theory is quite a bit clear to me, but the jargon when used again and again in the explanation becomes tedious to follow up. Due to this, forget about writing the code for the equations involved, even understanding the role of each variable in the theory is quite unclear sometimes.
So, what I am looking for is to be able to understand how I can implement Kalman Filters using the 2 functions - predict and update - and also use the matrices and parameters involved in the equations provided I have the center point of the object(aruco) I am tracking. The language I am comfortable in understanding is Python. Thanks!