Note:
- All measurements are taken/derived from 2D coordinates based on the video camera.
- Arm is a shorthand for angle between the hip to the armpit to the elbow. For all intents and purposes, arm can be substituted for some other angle in the context of this question.
Setup A
Suppose I have a person standing infront of a video camera and lifts up his arm such that it is 90 degrees.
If he moves back a 100 steps, and maintains the same stance, that arm will still measure 90 degrees.
The camera is not moving.
Setup B
Suppose the same person instead moves diagonally a 100 steps away from the original position, the camera tries its best to follow the movement of the person towards the destination.
The angle of the arm will now measure 60 degrees relative to the camera.
Setup C
Suppose the same person instead moves rightward a 100 steps away from the original position, the camera tries its best to follow the movement of the person towards the destination.
The angle of the arm will now measure 70 degrees relative to the camera.
My question is:
How do I accurately account for angular change given my only input is the 2D picture/video frames generated by the camera?
Meaning I would like to know that the 90 degrees of the arm is actually still 90 degrees instead of the 60/70 degrees that my video camera is seeing.
Is this something that is possible to do given only that data?