A good open source Photogrammetry software (Bundler/PMVS2 license will be a problem, so we should either use IGN MicMac or improve Libmv)
Find the best surface from point cloud reconstruction algorithm (For e.g Poisson algo can be found in MeshLab) that also creates the textures and their UV with photo projection on the generated surface .
1)+2)= 3) Create a all-in-one software where you input the photo and output the generated textured mesh. (If you want people to collaborate don’t expect them to type command line in a terminal and import/export in Meshlab, we must provide an all in one gui, It could even be a web app like the ID editor of OpenSteetMap.org)
A website where people can track the progress and see the world 3D reconstruction (which area have already been flown over and reconstructed in 3D ? which area haven’t yet ?)
Find who could sponsor and provide server to store the photo and 3D world. A kickstarter campaign to raise people attention and have some money could be a great idea.
However there are some issues with a collaborative model for 3D map like that :
-As the photo are not taken in one shot, but instead everyone submit a photo set of his area, then the exposition, brightness and shadow direction will vary which will lead (maybe) to visible edges between the different contributed area.
The positioning of the 3D generated mesh is tricky : how to make sure that it fits OSM street data and that the border perfectly fit with the neighboring area ?
The advantage over HERE map or Google map is that even small villages could be reconstructed as long as someone has a drone there and takes few hours to fly over it.