Pattern-Matching CamPhone Images of City Buildings to Determine Location
“Two researchers at the University of Cambridge, UK are currently working on a brand new technology that will allow the owners of cameraphones to easily make their way through an unknown city.
Roberto Cipolla and Duncan Robertson have developed a program that can match a photograph of a building, taken by a low-res camera integrated into many modern cell phones, to a database of images on a remote server. Containing a three-dimensional model of the street, the database can work out precisely where you are and send back directions to help you get to your destination.
Unlike GPS receivers and positioning using cell phone base stations, the new technology cannot be shielded, has a precision of one metre, and can also tell which direction you are facing.
When a new photograph arrives, the system starts identifying vertical and horizontal edges. Next, the image is distorted so that it looks as though the photo was taken face-on. The software then locates key points, such as the corners of buildings, windows and doors, and looks through the database for matching data, using the positioning info from the nearest cell phone base station as a guide.
At the moment the researchers are building a prototype to cover all the buildings in Cambridge city center. However, it is not known yet whether the system will be commercially available in the end.”