Surface Prediction for Spatial Augmented Reality

摘要

Image projection in spatial augmented reality requires tracking of non-rigid surfaces to be effective. When a surface is moving quickly, simply using the measured deformation of the surface may not be adequate as projectors often suffer from lag and timing delays. This paper uses a novel approach for predicting the motion of a non-rigid surface so images can be projected ahead of time to compensate for any delays. The extended Kalman filter based algorithm is evaluated using an experimental setup where an image is project onto a deformable surface being perturbed by "random" forces. The results are quite positive, showing a visible improvement over using standard projection techniques. Additionally, the error results show that the algorithm can be used in most surface tracking applications.