If a machine can drive a car, I believe it would not be long that smartphones could serve as line and service judges, especially with the new 115 cm fixed height rule. For those of you who are programmers, I believe OpenCV is the way to go. https://en.wikipedia.org/wiki/OpenCV https://opencv.org/ I found some books which maybe helpful to learn about OpenCV, which you can download here: http://gen.lib.rus.ec/search.php?&req=opencv&phrase=0&view=simple&column=def&sort=year&sortmode=DESC An interesting youtube channel on the basics of OpenCV is here Let us know if you are doing a project of similar nature and share with us your experience and questions.
Before I open it and potentially waste my time on a wild goose chase I ask: A) What does openCV do that makes it advantageous? B) and can you offer your own personal experience first so that people can give a more specific answer?
Firstly let me make at absolutely clear that I come from the good old dark ages where pen and paper were the norm and that I have a daily struggle with any sort of techy software. so please bear with me Having got that out of the way, can somebody explain to me just how using smartphones would work as line-judging devices, are you suggesting that they be run basically as cameras, placing them on each line? and that any line call challenge be reviewed as per the present hawkeye system. If so that by my calculation works out at 16 phones per court, 80 for something like the AE. Admittedly hawkeye is getting faster but do we really want to introduce an extra delay into the game whilst the software does its work
Yes, high fps are needed. BWF's hawkeyes shoot at 660 fps, I think. I did some research and found that Sony Xperia XZ Premium can shoot at 960 fps. They are selling it on Amazon for US$599. That is kinda expensive; hopefully the price of slow motion smartphones would come down in 2 or 3 years. [Correction: Xperia XZs can also shoot at 960 fps, and Amazon is selling it for US$399 only. Sorry.] There are a number of phones that can shoot at 480 fps, which are cheaper. But are they good enough for badminton?
That's a question of good engineering! HawkEye is not automatic yet, there is a human operator, and there used to be(?) a human umpire in the process. In principle, there's no reason why the system could not be fully automatic and decide and relay the decision within 100ms, faster than any human could.
Slow-motion clips with this device are limited to 6 seconds though. Maybe this is a limitation of the storage, but it doesn't bode well. The best camera I have here is an ipad Pro, which is not very good. Nevertheless, I'll try to gather example video (something I've planned for a long time) to see whether a certain FPS count could be feasible at all.
That's because of recording at 1080p. If you choose a lower resolution like 480p, you'll use up less bandwidth and memory, thus increasing the storage time. I believe the upcoming Samsung S9 will have 1000 fps recording
http://www.cnn.com/2017/03/21/tennis/tennis-tech-gadget/index.html I'd be very interested in how this In/Out tennis system can be adapted for badminton.
As I know OpenCV has a built-in function for contouring an image. Just in case, maybe @phihag interested in some PDF documents : • Shuttlecock Training and Tracjectory Estimation using Microsoft Kinect sensor • Flight Trajectory Simulation of Badminton Shuttlecocks
the Samsung Galaxy S9 Plus has dual cameras!! It can perform dual video recordings!! That's gotta be better than BWF's Hawkeye!!
thanks, Master. The Microsoft Kinect project also uses OpenCV, which I think is the standard library for this type of project. These researchers are based in Pune. The References in the Pune paper are very useful. It seems there are quite a bit of people writing codes in OpenCV for badminton. This guy from Guangxi uses background subtraction and morphological opening for badminton hawkeye! https://www.atlantis-press.com/proceedings/isrme-15/18589 This guy from Denmark has done some good work as well. http://www2.imm.dtu.dk/pubdb/views/edoc_download.php/6575/pdf/imm6575.pdf
@phihag I found a document coming from Technical University of Denmark Title: Badminton shot classification in compressed video with baseline angled camera Inside that document, you will find about shuttlecock dynamics (2D & 3D model) and also a little thing about OpenCV. Link : http://www2.imm.dtu.dk/pubdb/views/edoc_download.php/6575/pdf/imm6575.pdf
I believe that link has already been posted by @pcl99, one post above yours . To me, the image processing parts look interesting, although not spectacular. Especially when the video of the shuttle touching the ground is sent to the umpire, we can do with a relatively stupid analysis though, provided we can isolate the seconds when the shuttle touches the ground.
Yeah, I didn't realize at the time I posted it. Yes, image processing is going to the limit and it is used a lot in badminton analysis. By image processing, the analysis made on every aspect of badminton. The moving of player, the shuttlecock movement, the repetition movement, the racket movement, a detailed motion pattern, and so on. Badminton should really jump further into these high tech things.
for service fault call, the app needs to track (a) the shuttlecock, (b) the racket, (c) both hands, and (d) movement of the player generally. Hope that's not asking too much for a US$399 smartphone. Fortunately, I think 120 fps or 240fps is ok for service fault call as the shuttle doesn't move very fast during service. I am currently reading Chapter 4 "Controlling a Phone App with Your Suave Gestures" of OpenCV for Secret Agents (by Joseph Howse) for ideas. for line call, i think the $399 Sony Xperia XZs should be ok as it has a 960 fps camera. I don't think the task is computationally demanding even for a smartphone. For that, I am reading Android Application Programming with OpenCV 3: Build Android apps to capture, manipulate, and track objects in 2D and 3D by Joseph Howse. Good stuff.
