pixel to pixel with the trajectory of monocular visual odometry

Question

i have a trajectory computeted with a monocular visual odometry algorithm (scale is not known). In my first image I marked the middle of an object P(x,y). Is it possible to project the middle of the object (the Pixel P(x,y)) into the other frames by using the trajectory with unknown scale?

with known scale I could used the opencv function "cv2.projectPoints", but is it possible without scale?

-depth1-3 is unkown

-P1(x1,y1) is known

-trajectory is known (without scale)

-P2 and P3 is the result i would like to get