This odd-looking ball can automatically take a panoramic image whenever you throw it up into the air. Seriously, that’s then entire set of operating instructions for the device. Inside, a 3D printed frame hosts an array of 36 cellphone cameras, each capable of taking a two megapixel image. Also included is an accelerometer. When it senses the change in momentum associated with the apex of its vertical trajectory it snaps an image with all of the cameras at the same time. The result is a spherical image with no obstructions-like a tripod or other support mechanism. The images are automatically stitched together and displayed on a computer which allows the user to pan and zoom.
The whole story is told in the video after the break. The example images shown are quite good, although there are a few artifacts where the segments meet. Most notably, color variances between the images captured by different CCD modules. We’d image that this can be fixed automatically in software if a talented programmer were willing to put in the time. The thing about spherical photos is that methods using post processing to unwrap an image always have some distortion to them. With that in mind, we think the ball camera is as good a solution as we’ve seen.
[Thanks Gregory and Hans]
72 thoughts on “Panoramic Ball Camera; Toss To Snap A Picture”
It should be named religion ball. Photographers look like praying on all photos if they want to catch camera before it falls to ground. Otherwise – it’s amazing concept, so innovative I can’t even start thinking how anyone stumbled upon this concept.
@yetihehe: I read somewhere that the retail version even ships with a 1000ml jug of superglue!
For real though, I think this is a great idea!
Really really great concept. I like it and i want one!
As @yetihehe says, gotta be sure you catch it.
From the movie it looks like there is some difference in exposure and white balance amongst the pictures, maybe that should be fixed.
Guess it would be easyer to take the images raw so you can change light, whitebalance and whatever on the PC after taking the pic.
This device is fukingly awesome
Attach some kind of strechy string to it and you can throw it beyond the edge of buildings/cliffs/whatever. Fucking sweet!
To be clear, accelerometer cannot detect the apex of the flight path directly, well maybe it could if the accelerometer was very accurate to take wind resistance in account. Instead, the accelerometer is used to measure acceleration, and the apex can be integrated from the acceleration data.
Put a rocket under it and a parachute on top and sell it to the military. Recoverable 360 degree surveillance for observing dug in enemies.
except that it will land in the enemy’s dug out!
I wonder how the special viewer works… If it maps all the images onto some polys in an OpenGL environment, I’m surprised there’s so much distortion…
But what about the unavoidable motion blur? Just the other day I read about motion-deblurring in some other blog:
If their accelerometer and CCD sampling circuitry is fast enough they can take the picture at the apex of the throw, where there is no motion.
you’re forgetting spin
Bet you $10,000.00US you cant throw that ball without any spin.
Not true. There is no *vertical* motion, but there is horizontal motion (and spin).
Install a small battery powered gyroscope to control spin while airborne.
I’ve been thinking about this when i saw the panoramatic mirror photos being unwrapped by gimp, but i wasn’t crazy enough to think that it can be accepted as good idea by other people :-)
It should be noted that there is already a product of this, but I think it was discontinued, wasn’t a huge market.
dont worry, it didnt stop this group from attaching “Patent pending” at the end of the video …
How does it sense the apogee? On-board accelerometers should show zero g as soon as the ball leaves the thrower’s hands and remain at zero until it is caught. Are the accelerometers sensitive enough to detect the effect of air resistance?
you mean it should show 9.81 g , the sensors do sense gravity
Actually, it’s 9.81 m.s^-2 or 1g
Something in free fall has constant acceleration, so you can’t detect the apex of the ball’s flight with an accelerometer. What this device actually does is integrate the acceleration when the ball is being thrown (before it is released) to get its initial velocity, which is then used to calculate the time to the apex.
Of course you can – the first instant that the accelerometer reads zero is the apex. It doesn’t matter that it’ll continue to read close to zero for the rest of the trajectory.
No, Pelrun, Matt is right. The acceleration is zero from the time it leaves the hand until it hits the ground. (Barring negligible air resistance.)
-Physics professor. There are several of us who follow Hack-a-Day. :-)
Sorry — should have said MEASURED acceleration is zero. The acceleration is 9.8 m/s/s, of course, but in free-fall the accelerometers measure zero.
that is not how acceleration works – you are thinking of velocity. The z-axis acceleration would read zero from after release until it lands. It would not change during flight, Matt is correct.
Jim – MS Physics, NIU.
No Matt is correct, acceleration is constant during free fall, and that includes the trajectory when the ball is moving upwards, the apex, and the downwards part. There must be some other system used to determine the apex.
Ignoring air resistance, the ball is in free-fall (constant acceleration) the *entire time*. The person you are responding to probably read the page this post links to:
“Our camera contains an accelerometer which we use to measure launch acceleration. Integration lets us predict rise time to the highest point, where we trigger the exposure.”
No, you can’t. The same instant that the ball leaves your hands, it is in free fall, and It accelerometers should read exactly 0 until it is catched. For this size and speed, I think you can ignore friction.
The Acceleration time graph would have a positive spike at the beginning from the user’s throw but once it leaves the users hand it accelerates due to gravity at a constant -9.81ms^2 nothing happens to the acceleration at the peak of the flight. As matt said the accelerometer is used to gather data about the throw which is then extrapolated to work out the height.
Integrate acceleration, and you get speed+constant. If you know the speed at some prior point in time, for example if you can determine when the device is at rest, then you can detect the apex.
It’s actually a bit more difficult since there is not only vertical speed but also horizontal and spin to take into account, but that’s the idea.
Would someone please just link their Wii controller to their PC and throw it (the Wiimote) in the air to settle this? :)
I actually did that before I built this prototype ;)
I really doubt it uses CCD’s.
the imagers are cmos based
Just what kind of acceleration can this thing take, because I was thinking of somthing with a little more thrust than a small woman. http://en.wikipedia.org/wiki/Project_HARP
Just keep an eye out for mossad ay? ;)
I like the way you think.
Yeah, no obstructions except that stupid looking photographer. :)
So here comes the idea: add a parachute and launch device (no rockets needed) with remote control (could be sold separatelly as accessory for the camera).
Very cool anyway.
Should use Hugin to do the stitching – it removes vignettes
If there was enough room inside would some kind of manually spun gyroscope stop it from spinning and causing blur? I say manual because then you don’t have to worry about motors.
A gyroscope would certainly help, two would probably be better.
Making it manual is a brilliant idea.
I am uncertain about how effective this would be. I DO however remember a project that used an accelerometer to capture the required values for debluring an image. Add more sensors and you could do the same here.
To those suggesting military applications. There are several different kinds of cameras like this designed for military use, there are throwable ones and ones launched using standard grenade launchers.
The grenade launcher ones that I am aware of can be launched and transmit video for about 30 seconds. These are launched along with star parachute rounds at night.
this thing would have useful military applications. you could stay under cover and get a birds eye view of the combat situation. like looking for snipers while in a foxhole.
its a kino version 1
I can’t wait until V2 comes around. I would love to be able to fly my kino around manually.
What camera module does it use?
“36 fixed-focus 2 megapixel mobile phone camera modules”
made by ST, probly something like VS6724 / VS6725
hah I guessed correctly :)
ATtiny24, AVR UC3B, STM VS6724, […]”
VS6724 has buildin jpeg dsp, all you have to do is tell it “snap me a picture”, then you just download that picture at the speed you decide, so even low power AVR can do it.
Not true unfortunately, you need to be able to capture the images as they are captured by the imager. So AVR does not work ;)
I am quite impressed with this approach and am glad to see so much interest in this type of project. Readers interested in wide field of view high resolution imaging might check out this pdf. Note the spec’s at the bottom of page 7.
this.. is.. awesome!
Easily one of the coolest things I’ve seen all year. Useful in so many different ways for so many different people.
These would make collecting documentation a dream, since you could send a luddite out with a ball and tell him to toss it into the air while walking around an area of concern.
This is great…they need to make a video version and make it super-strong, then use it as the ball in Soccer and Baseball…it might actually make those sports worth watching, on TV anyway. Heck, it could even make Tennis interesting…for a while.
How is the camera triggering done? When the ball is thrown it is essentially in free fall until catched. Is air drag sufficient to determine apogee? I doubt the two uc’s have enough computing power to do it with the cameras.
Should have read it more carefully..”When it senses the change in momentum associated with the apex of its vertical trajectory it snaps an image”
Anyway, how does this work?
My layman’s understanding of the iPhone leads me to believe that the accelerometer is essentially the same principle here.
The accelerometer will detect the apogee through the namesake change in acceleration. More specifically the lack thereof, for a moment.
Having seen such concise answers on the subject, I wish I could delete my comment and not feel like such an idiot.
Calculus, I would guess. It knows only it’s acceleration, but it knows it’s acceleration from time 0 through the flight. Knowing how fast it is going at time 0, and integrating it’s acceleration function (which should be just gravity) it can determine velocity. Knowing it’s velocity, it can tell when it’s at an apex.
Less theoretical, just figure out how fast the ball is thrown and at what angle to the ground. Take that vector, and figure out just the upward movement. When time*gravity=initial_upward_vector, you’ve hit apex. Drag will play a small part, you could add a drag calculation into the first part, but why? I few milliseconds error isn’t going to be more motion blur than the spin of the ball will be.
Does anyone actually know how accelerometers work?
The original concept developed for torpedoes is a free WEIGHT with a spring fore and aft, allowing the weight to move back when first accelerating and for’d when acceleration ceases (making the connection which set off the payload).
Awesome awesome awesome.
The software Hugin might be useful for stitching, blending and exposure balancing. Might be nice if the white balance for all of the cameras can be synchronized.
I’d love to have one of these. It’d be really useful for planetarium production.
This could be very useful for military applications. Imagine being pinned in a firefight and throwing one of these up instantly knowing the situation in 360 degrees from about 30 feet up.
Or toss it into a building before anyone has to stick their head into the doorway…
Could be an awesome military item to just launch it up in the air and see if what is near by. Find where shots are fired ect.
make it smaller, make it capable of transmitting live video to a receiver, add a parachute and come up with a launcher to get it up higher, and the military would want it more.
The distortion has nothing to do with stitching software, but everything with projecting a 3D world onto a 2D photograph.
The reason is doesn’t show with a normal camera is that it projects only a small portion of that world ‘around you’ onto the plane.
Photographing a bigger part, thorugh multiple cameras, moving lenses, wide angle lenses, or fish eye lenses, always leads to distortion.
this would be bad ass for making ambient occlusion diffusion maps.
I want to throw this ball with a catapult. would be awesome!
I see a lot of application for this in security related organizations and with adventurers..
Catapult or a slingshot to throw it up would be awesome..
Of military issues. Why not hook a Wi-Fi or Sat link and drop it from an unmanned aircraft as an observation platform. You could confirm your kills with it while exploding it. Imagine if Bin Laden saw that fly into his bunker window, picked it up thinking it was a UFO and then 7 guys burst while he’s busy watching a “Western” on TV. Fun stuff. Or put mirrors on it and drop it inside the NYC New Years ball. Imagine the possible.
Please be kind and respectful to help make the comments section excellent. (Comment Policy)