A little while back I saw a video on YouTube on something called Photosynth from Microsoft. In the demonstration, multiple photographs of the same object were stitched together in a 3D space. Quite nifty. The software seemed to know where each camera was in relation to the object and to others.
If even a small percentage of those photos had GPS data available, it shouldn't be too difficult to figure out the GPS data for other photos. This opens up the possibility of a new service. Lost in some city and you're without GPS support? Take a picture with your cellphone and send it off to some service that uses some of the Photosynth technology to figure out where you are. Going through your pictures after a long trip but don't remember where some of them were taken? Upload them to the service and let it figure out the location for you. This can tie in with other services such as mapping and directions, and the more people use it, the better it gets.
Unfortunately, I think the Photosynth software requires Windows, but if that requirement is lifted or the technology is released to the public, alot of fun can be had.
I agree
No opinion
I disagree
Add your comment
This is actually a novel concept. I have worked with robotics and this is done to provide mapping for the robot. If a robot is dropped into an unknown space, it maps out the area using a similar technique.
By providing a GPS reference occasionally the robot can average out the GPS reference points coincident with its image references, and greatly improve its location coordinates. Current GPS is reliable to about +/- 25 feet, so a small robot of 12 inches or less can get lost within that space. So image analysis provides dead reckoning and cross references to increase the accuracy to within +/- 6 inches or less. Well within the accuracy needed by robots of that size without having to purchase WAAS subscriptions or higher cost dGPS systems.
So this is a proven technology in research. Expanding this to correlate random photographs would probably need some form of compression technology. The photographs in raw form represent large amounts of information that needs to be stored and manipulated. Color distributions and variance, edge detection and variance, ... I do not have sufficient expertise to speak to the details of image processing for compressing 2D images for 3D representation. It certainly is done.
This might be useful for crime forensics. A picture of a kidnapping victim is sent while asking for ransom. The picture is automatically processed to provide the location of the victim because someone had taken a picture of that room previously and it is in a national database. A certain crack is seen in the wall, the window case has a certain imperfection, the color distribution on the wall, the shadows cast by a ceiling light, ...
This is doable, and from a national security perspective, desireable. In addition to providing geolocation based upon pictures, the location of people can continuously be monitored. I personally don't care if the government knows where I am every second of every day. However, terrorist and criminals would find it disturbing. Whenever there is a crime, the criminal and all those involved could quickly be rounded up through all the cameras mounted in shops and on the streets.
Good idea, see DARPA for a similar systems that have been proposed to provide for national security.
http://www.darpa.mil/body/news/2005/index.asp
This same technology allows a robot with an initial GPS reference before entering a building, to map out the entire building and transmit that information to people outside the building or around the world. This is useful because the GPS signals are weak and are often not available inside of buildings.
For military applications, the small robot could identify a specific person and transmit to the extraction team the internal construction of the facility. Small robots the size of a roach are developed for research purposes. Actual insects have been fitted with electronic instrumentation to remotely control them like you would a remote control toy car. Your camera image concept could further make these small bugs useful to intelligence agencies.