Andrew Gee
Research Interests: Summary
BiographyAndrew Gee is an Equator funded PhD student in the Real-Time Vision and Mobile and Wearable Computing research groups at Bristol University. His research is in the area of real-time, vision-based simultaneous localisation and mapping (SLAM) for ultra-mobile augmented reality (AR) using a handheld/wearable camera. Initial work has been to investigate the use of line segments to build a physically meaningful map of the environment whilst robustly tracking the motion of a camera. Project abstractAugmenting video sequences with virtual 3-D graphics and animation in a manner that they appear to be part of the original scene has a wide range of applications, from movie making to wearable computers. In this project we are interested in developing a robust and scalable augmented reality (AR) system which is capable of over-laying content on live video obtained from a wearable video camera. A fundamental aspect of AR systems is determining the 3-D relative motion of the viewing camera and the relative position of key points and surfaces in the scene. This is a particularly challenging task to achieve in real-time and when dealing with the erratic and wide area motion observed in wearable applications. Recent advances in visual structure from motion (SFM) and simultaneous location and mapping (SLAM) systems have demonstrated good real-time performance within limited environments provided that key features in a scene can be reliably extracted and tracked in real-time. The objective of this project will be to build on this work and to develop effective methods for identifying and tracking 3-D scene features in real-time and over wide areas, such as within a single room or within a small building. This will then be incorporated within a robust tracking framework based on the fusion of the visual data with other active sensors, such as gyroscopes and accelerometers, and a wide-area ultra-sound location system. |