I please request input regarding whether the following can be accomplished with today’s computer vision technology (or the like), and if it can I am interested in any estimates for performing the following.
The attached document links to three short, two-dimensional videos of a golfer moving a golf club during what is termed the pre-swing period. Just a specific section of the club is zoomed in on, largely because the movement sought to be analyzed can be so subtle that it can barely be seen let alone analyzed if the entire golf club were included in the frame.
Pre-measured, known dimensions of the golf club that can be input into an appropriate software analysis program include the total length of the club and the distances from its grip end to the centers of each of the florescent tape markers.
The object is to precisely track the markers (equating that to the movement of the entire club) and subsequently feed that data into an appropriate algorithm(s) to calculate a rotation point location occurring somewhere along the length of the club. The relevant plane of movement is the gravity plane or the one along which a sensation of clubhead weight or effective clubhead weight can be sensed by a golfer as swing is about to begin.
Different golfers can have rotation points in substantially different locations along the length of the club, even when the exact same club is used to test with. The motions shown are brief representations of what in the end can vary in speed, direction, and/or amount of displacement as much as golfers’ swings and characteristics themselves.
A close examination will reveal that the club movement is not purely rotational in nature and that simultaneously there is some generalized moving around of the club as a whole. This is typically the case among most golfers and obviously adds complexity to any developed algorithm(s). But with that said, very small segments of the movements will commonly be isolated and analyzed (sometimes compared with another corresponding segment[s] of the period), and to that end it might be possible for a final solution to ignore the overall generalized club movement. This is not firmly known yet, but an early version could almost certainly ignore this generalized club movement (translation?) until seen what is gotten from a cruder attempt first.
The videos were taken with an old style analog video recorder operating at best 60fps and maybe even less than that. Videos 1 & 2 comprise much slower and subtle movements and there should be little if any blur if extracting individual frames. Video 3, however, is a much faster and abrupt movement that is actually quite common for a large number of golfers (it can be even faster and much shorter in time and amount of club movement), and I have noted considerable blurring when someone else sent me a few individual frames of that video.
Assuming that the slower videos can be acceptably analyzed first to gain a consistently accurate solution (ideally calculating a rotation point location to within one-quarter of an inch or better if possible), I then need to learn whether today’s technology is somehow able to analyze video 3 in an equally efficient manner. If not, I need to try to get a better idea of what camera frame rate and/or resolution might be required in order for the application to be successful. And I also need to determine whether such cameras are already routinely in the possession of individual golfers (perhaps the biggest market for the application) or can be provided at a cost they can typically afford (not what a commercial clubfitting establishment might be able to afford).
I need to quickly look into this or any other possible technology (lasers, magnetic waves, etc.) that can precisely gain the needed solution before another attempt at an inertial sensor type of system, a route that has continually failed for more than a half-dozen years now (though the technology is getting a bit better). Thank you.
7 freelancers are bidding on average $23/hour for this job
I took a breath and said “I can’t say that I’m the best person for the job, because I haven’t met and won’t meet the other applicants. I can tell you why I think I’m qualified, if that would be helpful.”