We experimented with various configurations considering
the robustness and the speed of the system. The results show
that our method is applicable for real-time applications with
sufficient accuracy to be a user interface alternative to fiducial
marker tracking.
The experiments were performed on a small laptop computer with a 1.8GHz CPU, using a USB 2.0 camera with
640 × 480 resolution. These specs are in line with currently
available ultra mobile PC and high-end wearable computing
platforms. Intrinsic camera parameters were measured by
the calibration method from [33] using the implementation
of OpenCV [13] with a 6 × 8 checkerboard pattern. We used
ARTag [8] to compare marker-based camera pose estimation
with our fingertip tracking approach.