Instantaneous 3D motion from image derivatives using the Least Trimmed Square regression
Pattern Recognition Letters
Computer Vision and Image Understanding
TrackSense: infrastructure free precise indoor positioning using projected patterns
PERVASIVE'07 Proceedings of the 5th international conference on Pervasive computing
Markerless augmented reality for robotic helicoptor applications
RobVis'08 Proceedings of the 2nd international conference on Robot vision
AR-Room: a rapid prototyping framework for augmented reality applications
Multimedia Tools and Applications
Hi-index | 0.00 |
We address the problem of tracking the 3D position andorientation of a camera, using the images it acquires whilemoving freely in unmodeled, arbitrary environments. Thistask has a broad spectrum of useful applications in domainssuch as augmented reality and video post production. Mostof the existing methods for vision-based camera trackingare designed to operate in a batch, off-line mode, assumingthat the whole video sequence to be tracked is availablebefore tracking commences. Typically, such methodsoperate non-causally, processing video frames backwardsand forwards in time as they see fit. Furthermore, they resortto optimization in very high dimensional spaces, a processthat is computationally intensive. For these reasons,batch methods are inapplicable to tracking in on-line, time-criticalapplications such as video see-through augmentedreality. This paper puts forward a novel feature-based approachfor camera tracking. The proposed approach operateson images continuously as they are acquired, hasrealistic computational requirements and does not requiremodifications of the environment. Sample experimental resultsdemonstrating the feasibility of the approach on videoimages are also provided.