[Koppany Horvarth] set out to create a dirt-cheap optical tracking rig for VR that uses only two cameras and a certain amount of math to do its thing. He knew he could do theoretically, and wouldn’t cost a lot of money, but still required a lot of work and slightly absurd amount of math.
While playing around with a webcam that he’d set up to run an object-tracking Python script and discovered that his setup tended to display a translucent object with a LED inside of it as pure, washed-out white. This gave [Koppany] the idea that he could use such a light as part of his object tracking project. He 3D-printed 50mm hollow spheres out of transparent PLA, illuminated via a LED and powered by a 5V power supply hacked from an old USB cable. After dealing with some lens flares, he sanded down the PLA a little to diffuse the light and it worked like a charm.
To learn more check out his GitHub code repository. You can also take inspiration in some of the other motion tracking posts we’ve published in the past, like motion tracking on the cheap with a PIC and this OpenCV Airsoft turret.
What about ping-pong balls instead?
Too easy :p
nice actually, this is the way the PSmove is working and there are open source libs that use it : https://github.com/cboulay/PSMoveService and https://github.com/thp/psmoveapi
the former you can also use to track light blobs for a VR headset f.ex.
how to make a ping pong ball that doesn’t bounce….
But seriously now, this is a really cool project, thanks for sharing.
There’s a reason why many (most?) visual tracking systems use WHITE markers on DARK clothing … because with enough lighting in the room you get quite distinctive “almost white” spots in your video footage from those. No need to actually power them, it’s the contrast that’s doing the magic in “standard setups”.
Sure – this project is a nice hack, but reproducing something that is usually done with the same goal, but way, way cheaper, is a kind of a “I hack it because I can”-hack. Which is fine. But expensive.
It seems that the creator either didn’t bother to go fetch a ping pong ball (or order them on amazon) … or just thought it would be easier to print one. Or maybe he knows that by not choosing the obvious route he would get spiked readers from hackaday to talk about his project, thus making him proud and getting attention on his project.
:thinking_emoji:
I didn’t have any ping pong balls around and didn’t feel like ordering and waiting for them to arrive. Plus, I’ve been trying to do projects without spending money and only using what I have on hand, and I had a 3D printed with clear filament. :)
Also, ping pong balls are a little small and I wanted the camera to pick it up from a nice distance, which is why I didn’t print the ball to be the same size as a ping pong ball. :p
Think about the professional motion tracking setups used in movies: reflective spheres illuminated by infrared leds around the camera lenses
Newsflash June 3rd 2022: Self driving cars now 100% reliable as new laws come into force requiring motorists to have 30cm illuminated ball on roof of car…..
They already require lights on cars I hear it whispered.
Also they are actually really talking about forcing cars to have things like RF communication to nearby cars, although there is some slight resistance against it fortunately. But I fear there is no future of driving without the big companies and the government knowing at all times where you are where you go (often in advance) and how fast you go. Future is bleak for those who cherish a modicum of privacy :/
Oh and with the 360° cameras on self-driving cars around you you’ll be observed at all times too.
Well yah, but they’re not standardised in size/shape or anything. I bet one system or other starts rearending those Cadillacs with with lights right out at the corners.
I’m happy amateurs are working on such VR projects because I really don’t like how the commercial side of things are all ‘only windows10′ creative edition’ and ‘you need this crap spyware to run it’ and all such shenanigans.
SteamVR runs on Linux.