Information Overlay for Camera Phones in Indoor Environments

Increasingly, cell phones are used to browse for information while location systems assist in gathering information that is most appropriate to the user’s current location. We seek to take this one step further and actually overlay information on to the physical world using the cell phone’s camera and thereby minimize a user’s cognitive effort. This “magic lens” approach has many applications of which we are exploring two: indoor building navigation and dynamic directory assistance. In essence, we match “landmarks” identified in the camera image with those stored in a building database. We use two different types of features – floor corners that can be matched against a floorplan and SIFT features that can be matched to a database constructed from other images. The camera’s pose can be determined exactly from a match and information can be properly aligned so that it can overlay directly onto the phone’s image display. In this paper, we present early results that demonstrate it is possible to realize this capability for a variety of indoor environments. Latency is shown to already be reasonable and likely to be improved by further optimizations. Our goal is to further explore the computational tradeoff between the server and phone client so as to achieve an acceptable latency of a few seconds.