How can i estimate the pixel coordinates?

I have a project that I need to follow the points like in the picture. I made a picture like this as a representation. I get the coordinates of the red dots as pixel coordinates once per second. Is there a way I can predict the green dots when I don’t take measurements afterwards? When I use the 2d-kalman filter (x,y,x’,y’ states) it keeps guessing in the direction of the blue arrow. I want to make correct predictions on the green dots, not this way. How can I do that.!


the arrow is a linear prediction. it uses two data points to extrapolate a line.

you need a higher order prediction.

that means you need to use more data points (stateless) or more state (kalman filter).

same question:

@crackwitz thanks for the answer. So, can I do this with deep learning? How should I search if I can?

what? no. what gave you that idea?