X-Ray Vision for Mobile Outdoor Augmented Reality

I completed my PhD in 2009 at the University of South Australia, working under the supervision of Dr Wayne Piekarski and Professor Bruce H. Thomas from the Wearable Computer Laboratory.

This dissertation presents a system for enabling the superman-like ability of x-ray vision outdoors. Looking through objects improves spatial awareness by allowing the viewer to see more of the environment than their immediate surroundings. Augmented reality is the registration of computer-generated graphics registered on a user’s view of the world. Using augmented reality technology, the appearance of a hidden location can be displayed on top of an occluding surface, emulating the effect of seeing through an object. Displaying this on the user’s display may be confusing or difficult to accurately comprehend. To overcome this, improved visualisations were developed that address issues that arise when generating x-ray vision with augmented reality.

This dissertation describes an x-ray vision system working in outdoor environments and presents a number of related contributions. The x-ray vision system uses image-based rendering techniques with geometric models and video streamed from occluded locations, to reconstruct a photorealistic view to present to the user. An evaluation is presented confirming augmented reality x-ray vision to be a viable method for viewing and comprehending remote locations. The visualisation of x-ray vision was compared with conventional techniques of watching video footage and found to have a number of advantages. Further visualisations and user interaction techniques are presented, which improve the user’s spatial perception when using x-ray vision. The visualisations include the edge overlay for improving the sense of depth, the tunnel cut-out for showing additional occluding layers and the two-handed magic lens technique to control the x-ray vision viewport. A number of side-effects of the photorealistic rendering are also demonstrated. The user can rewind time and view an environment as it appeared in the past and visible objects can vanish by drawing over the top of them.

A number of methods are presented for acquiring the images and models required for the photorealistic reconstruction. A mobile robot was constructed, which can be remotely navigated by the augmented reality system to capture remote video. Video can also be shared between multiple users with mobile augmented reality systems, or obtained directly from pre-recorded video files. Models are created using in-situ modelling techniques for unprepared environments, generated using semi-automated computer vision techniques, or read from pre-created files. This thesis also presents a novel compact design and construction for a wearable augmented reality computer allowing a user to be completely mobile to view x-ray visualisations. Unique visualisations, flexible data acquisition and complete mobility are combined, demonstrating x-ray vision to be a compelling application for mobile outdoor augmented reality.

A pdf of my thesis can be downloaded here: Link [21meg]