Detecting Visual Tags
This work addresses the detection by computer vision of visual tags. Examples of visual tags include color-coded tags, which encode identity. It is possible to generate thousands of distinct color codings for a badge of the type illustrated at top-left. Such tags could be worn by people or attached to objects, and are cheap and disposable. Other examples of visual tags are logos. The illustration at bottom-left shows the detection of a Coca-Cola logo. This detection could be used to add intelligence to a Coke vending machine, enabling the machine to issue a special greeting to a customer wearing the logo, or informing the customer of a discount in the price of a soft-drink. A further use of visual tags is to transmit information about the wearer - for example, to indicate that the wearer is infirm, elderly or handicapped, so that a detecting system such as an elevator can offer a special level of service.
Background & Objective: (a) Identity tags: cheap and disposable identity tags have many uses for tagging and tracking objects. (b) Logos: an interactive vending machine, which responds when a customer is wearing a product logo, could increase sales and could encourage customers to wear the logo thereby providing advertising. (c) Information tags: a badge indicating that the wearer is elderly or infirm could be used to automatically invoke special levels of service from the detecting system - for example, an elevator system could level the floor more accurately on arrival if a person in a wheelchair is present.
Technical Discussion: The approach is general enough to work for a variety of tag appearances and sizes. The system uses two cues - (a) shape of the tag, and (b) the internal design of the tag. Processing speed is about 1Hz on a standard PC processor, for a 240-x-240 pixel image. The system works for tag sizes in the image down to about 15 pixels diameter. Thus the system can operate to a range of about 10 feet for a 240-x-240 pixel image, and this range increases for higher image resolutions.
Technology Area: Computer Vision
Modification Date: June 13, 2008