As part of a “disruptive technologies” course at the United States Military Academy, [Roy D. Ragsdale] produced a working prototype of a Google Street View-like system called PhotoTrail. Like its corporate-backed inspiration, the system captures georeferenced 360-degree panoramas that can be viewed interactively in a web browser…but at a hardware cost of only around $300. [Ragsdale’s] prototype is based entirely on consumer-grade off-the-shelf components and open source software, all tied together by the yin and yang of DIY: foam core board and a few Python scripts.
This article from IEEE Spectrum magazine provides some background on the selection of parts and construction of the system, including a hardware shopping list and a list of links to all of the open source packages used.
The PhotoTrail prototype is surprisingly small and lightweight. A vehicle isn’t even required; the camera array can be carried overhead by a single person, making it possible to capture remote locations. But [Roy] expects future revisions to be even smaller and less obtrusive, perhaps mounted to a headband. Mount Everest awaits!
FIRST!!!
I wonder if using two panoramic mirrors and two cameras you could gather the same data, thus cutting the cost by a decent amount.
FIsrt!!
Heh, I read “mounted on a headboard” at first…..
@dan
Third!!
Augmented reality from this onto a windshield screen anyone?
The biggest hack is buying 1600×1190 camera for $25, even if its just a piece of crap made by M$ that needs a searchlight in your face to be usable as a webcam.
His setup needs 8 seconds per picture and cant sync up all the cameras so its unusable while moving. Would be a lot easier to just buy couple of ordinary consumer digital cameras with remote shutter and sync everything up later from SD cards.
@rasz,
While it’s true that 8 seconds is quite slow, even at the modest speed of 30 mph that only puts about 350 feet between captures.
From what I understood he cant take ANY pictures while moving because cameras constantly take pictures during this 8 seconds and software later picks the best ones (focus/light). This might be a direct effect of choosing one of the crappiest webcam vendor possible :).
M$ Live search something blabla meetoo technology is powered by EveryScape. They use Nikon D70 with Sigma 8mm F3.5 EX DG. Thats ~$800+$370 times four = $4700 for a perfect setup.
You could do fine with four used $50 Nikon Coolpix cameras and hacked peephole fisheye lenses. Put them in fast shutter mode and you can race while taking pictures.
Get 4-8 of the Flip Video UltraHD cams, $250 each new, much cheaper used, 2 hours of HD video on solid-state memory with built-in USB. You could probably rig some software to dump the video every x minutes and pick frames it likes for whatever features you want. As long as 30fps isn’t too slow it should be fine at most speeds.
once again the do’ers out think the thinkers. guys i am only scratching the surface of projects like this in my head but all of the idea presented here seem more reasonable then the approach taken.
one thing i do see as a sharp contrast is that each of your solutions would require substantially more software and hardware on the processing side. though i really doubt mr. sam wouldn’t shell it out
but i digress, anyone got a quote on the hardware to dump video, process the pics for the best and nearest result to each ‘tick’ and store enough of them to compose a decent sized hotzone (say we just do several city blocks per local system)
I would guess several million to develop the software side and if you were aloud to use off the shelf hardware about a grand or so for the truck mount unit and if you wanted to store iraq as it was when the last truck drove by another couple million on hard drives and networks
of course these numbers are for us doing the job, they will pay someone 10 times that for half he system then scrap it
ideally you would want a hardware solution, like the one google uses (ladybug2/ladybug2). $10,000 last year including software.
Really stop hating on the man. It’s a proof of concept.
What about a reflective dome with a single camera? We used to do that all the time for panoramas, Sure you get a lower overall resolution, but no stitching, fast results.
Lets see if I could be a average hackaday commenter… “I didn’t do that, but I could do it better, but I wont”
Is this the best you all can do? Where is my holodeck? 360 views are nice, but want a holodeck-style rig like the one on Star Trek. Can ANY of you geniuses pull something like that off?
Why can’t we use a rotating camera instead? I don’t see a problem at all if the shutter speed is quick enough and assuming photos are taken day light which provides a lot of lum.
once again the do’ers out think the thinkers. guys i am only scratching the surface of projects like this in my head but all of the idea presented here seem more reasonable then the approach taken.
Except that the guy actually did it, and everyone here is just talking smack.
The hardware is trivial. Any configuration that captures 360degrees is enough. The software to link up the images is a pain.
@mrgay its not trivial, as its the hardware that limits this hack to stationary captures. Hardware, not software.
@rasz – Throw more money into better cameras, and you solve the stationary capture issue. Trivial =/= cheap.
@Chris
>You could do fine with four used $50 Nikon
>Coolpix cameras and hacked peephole fisheye
>lenses. Put them in fast shutter mode and you
>can race while taking pictures.
how is that more money?
Does anyone know of any way or commercial available software to stitch video together. I have six wide angle video feeds taken at the same time and I want to stitch them into a 360 VR type video? Any advice or suggestions?
maybe this will help
http://panotools.sourceforge.net/