Crawl the web for images, use the GPS EXIF data to find clusters, and use image processing to stitch together 3D environments from the photos.
It would not be as good as streetview though.
It would not be as good as streetview though.
edit: believe it was related to http://en.wikipedia.org/wiki/Photosynth
And if you scripted a time limit for images, e.g. only within a few hours, you'd heavily limit the amount of usable images you had.
With "agents in the field" Google could potentially update its feeds whenever and as often as it needed. The difference would be having an in-house image-generating infrastructure or...well...tag-searching Instagram.
http://www.fastcodesign.com/1669688/googles-photo-tours-sugg...
http://google-latlong.blogspot.com/2012/04/visit-global-land...