I want to compare the activity of two virtual cameras moving around in different virtual scenes across few hundred samples. I want to be able to show and display which of the cameras have had more activity.

Each camera frame sample consists of position in the scene's world coordinates and local euler rotation in radians. E.g.

Camera 1

Frame 1: position x=30.5862 y=-1.5571 z=0.7617, rotation x=1.6968 y=-0.0049 z=-4.6078
Frame 2: position x=29.5252 y=-1.5496 z=0.7500, rotation x=1.5668 y=-0.0051 z=-4.6325

Camera 2

Frame 1: position x=2.0696 y=10.7080 z=5.4238, rotation x=1.0245 y=-0.1208 z=2.9924
Frame 2: position x=2.0693 y=10.7080 z=5.4239, rotation x=1.0245 y=-0.1208 z=2.9930

Scale of the scene is completely lost, so positional coordinate unit does not map to a known measure of length. E.g. change of 1.0 units along x coordinate might be one millimeter or one meter.


Given that the we are observing a camera here, even a small variation in cameras rotation will affect what the camera sees. What are the measurements I can produce out of these samples, that represent well the amount of activity each camera have had?

One point to consider, when thinking about the rotation of the camera, if the change happens around the camera's forward vector, the things the camera sees will not change that much, just rotate around. For example in this picture, whether the change happened around y-coordinate or z-coordinate will hugely affect what the camera has seen during the animation. We can assume the camera's forward, right and up vectors are known.

I have thought of calculating a vector about the difference in position between previous and current frame, but I am wondering whether they are comparable given that the scales of each scene are different. I have not yet come up a way to present rotation in clear and comparable way.

I have not yet found any material where positional and rotational samples are compared with this approach.

  • $\begingroup$ If, over the course of a year, I grow taller by one unit, and my friend grows heavier by one unit, which one of us changed more? The reason you don't see positional and rotation data compared is that you need to choose a units conversion to make that a reasonable thing to do. (It's peculiar because radians are "unitless", but it's still a problem.) $\endgroup$ Nov 4, 2020 at 13:09
  • $\begingroup$ There was a mistake in the question and I have edited it. We can assume that the x, y and z coordinates are units of length. And comparison should be done between the cameras, which had more changes in position and which had more changes in rotation. $\endgroup$
    – Toothery
    Nov 4, 2020 at 15:14
  • $\begingroup$ I stand by my comment. $\endgroup$ Nov 4, 2020 at 22:23


You must log in to answer this question.

Browse other questions tagged .