Amazing mapping tool in development

[evomedia]

Amazing mapping calibration work going... this guy is developing a 'spatially augmented reality toolkit', using light sensors to calibrate and map the projector

http://augmentedengineering.wordpres...ogress-videos/

YouTube - Re-implementation of Projector Calibration with Embedded Light Sensors

YouTube - Re-implementation of Projector Calibration with Embedded Light Sensors

[MoRpH]

Thats pretty cool, was only a matter of time really til someone combined a mapping routine/images with a camera to do this automatically.

[many2]

I fail to see the innovation - care to explain what's amazing about it ?

[MoRpH]

I believe its automatically doing the mapping to the object by comparing test grid projected with what a camera sees when pointed @ the projected lines/grid.

[bilderbuchi]

not precisely; there are sensors (photodiodes) embedded in the box (at the corners presumably) which calculate their position in the projector cone from the light/dark pattern sequence illuminating them.

[evomedia]

Yep that's how it works, why is that exciting? well essentially you could throw a screen up anywhere and the image could auto align to the screen,

If you had a mapping installation with multiple boxes, rather than having to take a pic and work back, or sit with a computer and set masks for each target, you could simply thrown the shapes up anywhere and the sensors could feed back the data they recieve and auto align the mapping data. You could practically get anyone to stack boxes or shapes anywhere within the projector field of view and the software could work out where each shape sits and warp the map for each object so its correct perspective wise.

All you would have to do is embed the sensors into each objects corners, tell the software how to interpret them and then you could put the objects in any configuration and they would auto map. Its projection mapping meets augmented reality, basically mapping without the hard work.

[MoRpH]

Hmmm having shit embedded in the box actually makes it less cool IMHO, cause then its not as portable.

I think my idea would be cooler and more portable. Have a camera next to your projector, project a flat grid, then have the image of the distorted projection go back to the software via the camera, then the software looks @ the distortion and figures out where objects are in 3D space and maps them accordingly.

Now thats cooler than sensors in boxes

[evomedia]

I totally agree morph, what I was also thinking, the fact that augmented reality could be coupled with mapping is what makes for some interesting opportunities,

I think the main reason for the sensors is the same reason for the glowing ball on the playstation move, they feed back actual points in space accurately, of course the same could be achieved by visual markers like a barcode thats already been implemented quiet heavility in augmented solutions.

So yep a system that projected a image with a camera to capture the pattern and have the computer auto align is more than possible, its expanding that to work with seperate shapes thats tricky, hence why putting a sensor into the shape itself gives a simple solution,

[bilderbuchi]

problem with AR markers is that the have to be big to be recognizable, and who wants big stickers on their objects/canvas/etc.
problem with the camera approach is that you need a fixed geometrical relation between the camera and the projector (i.e. bolt them together) to calculate the distortion correctly. on the other hand, they should be appreciably far apart to gain some parallaxis, otherwise you won't be able to detect the discontinuities in the pattern due to object edges.
maybe some really spiffy calibrating could nullify the fixed geometry issue, then it would be easier. you could also use something elegant like the bocode, but then you are back at putting stuff into your boxes again.

[MoRpH]

Originally Posted by evomedia

basically mapping without the hard work.

Setting up sensors in each object sound like alot of work to me.