Multiple View Motion Tracking of Gridded Surfaces using Topological Structure

AbstractThis paper describes a novel algorithm for interframe motion tracking of gridded surfaces across close range video sequences. The topological approach exploits the regular grid structure, simplifying feature extraction requirements, and reducing dimensionality of the projective tracking problem from eight to two. Multiple view grid asymmetry is implemented using loosely defined tracking fiducials that do not need to be explicitly detected or measured. Results from an experimental setup validate the accuracy and robustness of the algorithm. Timing results demonstrate throughput greater than the 15 frames per second camera data acquisition rate.