1

I assume that before you can measure a correct angle from a bodypart, you have to calibrate the camera? The camera must be set up perfect horizontal and vertikal in front of the object?
How do you do this? I can't find a topic about this.

2

I don't have much experience with this but here are some first thoughts.

A single video camera works with projected images.   It is exactly similar to how your eye works.  Therefore it is easiest to do some simple demonstrations with the eye to understand what is going on. 

Take a piece of stiff wire, such as cut from a coat hanger, or short piece of solid electrical house wire, pipe cleaner, etc. that will hold its shape.  Take a straight piece and bend an angle, ?, into it. (Simulates a person bending an elbow to ?, etc.) 

1)  hold the piece at arm’s length with one leg horizontal and rotate the angle slowly around.  The apparent angle seen ranges from ? as a maximum to 0°.   

2)  with one leg horizontal and ? at maximum (‘plane’ of the angle perpendicular to the line from your eye) now rotate the apex of the angle toward you.   The apparent angle will go from ? to 180°. 

Any angle ? then can range from 0 to 180° as seen depending on how the angle is oriented.

With one camera you have to be very selective in what you chose to measure & how you measure it. 

Golfer. Side view of a golfer swinging is the easiest as the planes of the swing and the sensor are parallel.

Sprinter. Another special case - again I don’t have experience doing this – would seem to be a sprinter running in a straight line, across the camera’s field of view and about the same distance from the camera on the right as the left (telephoto is best). The leg would be mostly vertical and if the camera were set up with the sensor plane vertical then I believe the angle of the leg would be pretty accurate.  If the legs move toward or away from the camera it would lose accuracy.

For motions with known orientations that are not parallel to the sensor plane I would try to develop a technique using calibration objects with angles to video before, after or during the sports video.  For example, video calibration angles could be made of PVC pipe with 45, 90° or 135°, etc. or better yet pieces of PVC pipe every 20 or 30°.   If a baseball batter swings up at 20° put the calibration angle in that plane and video it.  I don’t know if Kinovea has a tilt capability to match angular calibration for a technique like this.   

For other motions with unknown orientations as may occur in sports I don’t know what to suggest.  Maybe calibration angles could be put in after looking at the video.  For example, the batter was seen to swing up at 35° so video a calibration standard at 35°.

I believe these problems can best be dealt with using multi-camera systems with cameras viewing from different angles and computer processing of the video.

What application are you interested in?