Jul 7, 2011
Augmented reality involves the rendering of synthetic content into live video. The believable augmentation of reality requires that synthetic content interacts realistically with the real objects visible in this video. This means that real objects should occlude synthetic content which is at a greater distance, that synthetic labels should appear to adhere to real objects, and that synthetic objects should appear to bounce off the real environment where appropriate. In order to achieve these results the AR system needs to know what objects are visible, where they are relative to the camera, and what their shape is. This talk presents a range of approaches to this problem based on analysis of the incoming video stream, and particularly focusing on online reconstruction and recognition.
Prof. Anton van den Hengel
Australian Centre for Visual Technologies (ACVT), University of Adelaide
download the lecture presentation