Over the last few years cameras have moved from being an optional extra to standard equipment on Internet connected devices. Increased processing power on these same devices is creating possibilities for new kinds of interaction with Internet users. Using advanced computer science and mathematics applications can begin “seeing” through the camera: recognising objects and using this information to interpret input from users, or building a model of the environment in front of the camera and projecting virtual objects into the camera's field of view.
Read below for more information about a few of the experiments we've carried out in this area.
One of the fundamental problems in computer vision is recognising the boundaries of shapes in an image: for instance where does the background stop and a person's face start, or what is the outline of the next character on a sheet of paper? It's even more useful if we can detect straight edges, because they turn up in most applications and are easier to express mathematically in further processing. Our eventual aim in developing fundamental tools like edge detection is to be able to create more natural augmented reality applications that do not require markers or special props to work.
Motion detection is another fundamental problem in computer vision. In this experiment we have implemented a motion tracker that can track multiple centres of motion at the same time. Our next step is to build some gesture recognition logic.
Although it has been available in desktop and multimedia platforms for most of this decade, augmented reality has only come to broad attention in the last year or so with the implementation of some excellent toolkits for the Flash player platform. Like a “Smart Mirror", augmented reality applications let the user manipulate a three dimensional model by manipulating a prepared physical prop which the application can recognise using the computer's camera.
Almost all of the Augmented reality demonstrations in Flash use the excellent FLAR Toolkit. This example uses the recent released Alchemy branch of the FLAR Toolkit, which exhibits significantly better performance, with frame rates in the upper thirties.
you will need to print out the pdf marker.
