I just downloaded your software this morning and am very pleased at how it works.  It is the first sports software that I have used.

Recently I realized by reading discussions on the Tennis Talk website that shoulder upper arm rotation may play a much bigger part in the tennis serve than I had previously understood.  [Pronation is defined as lower arm rotation.]   I am now interested in measuring upper arm rotation and practicing it to improve my serve.

To illustrate here is a video of two of my serves. I put pieces of black plastic tape on the protruding elbow bones of my humerus to make the upper arm rotation stand out.   Casio FH100, 240fps, 1/10,000s. 


The tape moves suddenly indicating when I rotate my upper arm. Probably late? I can measure the motion linearly and assume a radius to the upper arm's rotation axis to get an angular rotation rate.

A similar upper arm rotation is supposed to be a significant contributor for baseball pitching.  For pitching I have read that the upper arm is capable of rotating up to 7000 degrees/sec, possibly one of the fastest rotation rates for the body parts.  The motion is powered by the stretched internal rotators, the pec and lat, large muscles.

I am not that knowledgeable in biomechanics so please double check my conclusions.

Anyway, if I could manually position a rotation axis in the upper arm to measure the position of a marker on the upper arm and repeat it on  subsequent frames that would be what I'm most interested in now.   The question is does such a tool offer any advantage over measuring the tape position linearly?   Is it possible to measure the rotation angle from a 2D video image where the axis also changes from one frame to the next?


(I can't comment on the mechanics of your serve or on tennis in general, hopefully someone more knowledgeable will chime in smile)

Indeed rotations are an extremely important part of almost all sports, so measuring angular speed would be super interesting in many contexts !

But as you point, the center of rotation is also moving, which makes things complicated.
Also, the rotation axis has to be exactly perpendicular to the image plane otherwise there is a bias due to perspective.

Technically… The first step would be to implement the "relative origin" for the tracking tool (coordinates of one tracked point expressed relatively to another tracked point).
Then we can probably figure out the maths to compute the angle change between successive images.

Speaking of baseball, rotations and angular speeds, the guys at DrivelineBaseball are apparently trying to hack a WiiMote to retrieve accurate rotation data. Sounds like a cool project !


I have not thought out very well or for very long.......dumb mistakes, etc.......so please don't take it otherwise and spend much time   

Let us say I place an axis as best I can in the upper arm.   The axis begins at the judged center of the elbow to runs to the judged ball on top of the humerus or other end for the axis that seems easiest to judge in the image.   There is a disc always centered on the axis and by eye you move it along the axis until its center is near some marker (the tape on my arm in my video, or bony protrusion or some other reference point on the body). By eye you make the disc perpendicular to the axis and intersecting a point on the marker. That gives one reference line for the angle, such as it is, for frame #1. 

Step to a later frame and place the same exact axis & disc image oriented as copied from the frame #1 into the later frame.   The orientation for the limb & its axis may have changed some.  I think that the center of the disc will always be at the same relative position along the axis (for example, 67% of the distance from the top to the bottom).  I'm not sure of this assumption for all camera conditions, wide angle, large angle limb change, etc., assume for now.  If you carry the image from the first frame into the later frame and reoriented it manually including the disc, would that do it?  The length change for the limb should help reorient & place the axis in the later frame. One line for the angle would be from frame #1 and the another from the later frame.   Orienting the axes seems similar to your perspective grid operations.

I have no idea of the errors in placing rotation axes into limbs in videos.   

FYI - general interest on rotation and a non-video technique.

Look at the internal rotation rates for lower, intermediate & advanced level players: