3 ms·
One way I see one could crowdsource the problem without turning this into a witch-hunt would be to collect as many pictures as possible and try to recreate a 3D
by AYBABTME 14y ago
One way I see one could crowdsource the problem without turning this into a witch-hunt would be to collect as many pictures as possible and try to recreate a 3D model of the scenes at various timestamps. Something like,
* Go on project website. There's a 3D mesh of the area.
* Upload your picture:
- Select a time frame you think fit your picture.
- Navigate to the location you took your picture from.
- Try to apply the picture on the mesh as well as possible.
* Or, look at what people posted:
- Try to correct other's time-positioning, or
- Try to correct other's spatial-positioning
- Fix the mesh for missing things.
* Investigators can use that to look for clues on the scene.
Like OpenMap or Google Maps where people can send corrections. Of course, I don't assume that this is trivial to implement. Just, an idea of how such future problems could be approached.
- scoot 14y agoMicrosoft's photosynth[1] sounds like a good fit to automate much of what you're describing (without the time function). [1]http://photosynth.net http://photosynth.net
- GrantS 14y agoIf anyone is attempting to build something like this, I'll just note that automating this process (spatio-temporally registering photographs and reconstructing 3D models of cities as they change over time) was essentially the topic of my PhD thesis, but on historical time scales (1860s-2000s) rather than minute-to-minute. Slides from my CVPR 2010 presentation here: http://www.cc.gatech.edu/~phlosoft/files/schindler10cvpr_slides.pdf http://www.cc.gatech.edu/~phlosoft/files/schindler10cvpr_sli... Videos: http://www.cc.gatech.edu/~phlosoft/ http://www.cc.gatech.edu/~phlosoft/ The best way to get a system running quickly would be to use Noah Snavely's Bundler project for the 3D reconstruction: http://phototour.cs.washington.edu/bundler/ http://phototour.cs.washington.edu/bundler/ and, at first, trust most EXIF tags for time.
- aosmith 14y agoSuper interesting, thanks for sharing!
- pilom 14y agoNorthrop Grumman makes a product which collects the images and automatically stitches them together to allow law enforcement to record where bad people come from before a bust. http://www.reuters.com/article/2012/08/30/us-northropgrumman-planes-idUSBRE87T02U20120830 http://www.reuters.com/article/2012/08/30/us-northropgrumman... If only this plane was in the air surveying all of Boston before the bombs went off <sarcasm?>