Virtual input using skin color model for robotic platform control

The aim of this research effort is to build a robust image-based virtual input for robotic platform control. Compared to the motion of the head, eye gaze, face or even the whole body, tracking of hand movement has become increasingly popular, efficient and more natural in human computer interaction. However available systems are invasive and require the user to wear gloves or markers. In this project, we propose a markerless hand tracking system as a virtual input device. We track the hand movement using skin color as detection cue because color allows fast processing, and is highly robust to geometric variations of hand pattern. The skin segmentation are modeled by a parametric skin modeling using single Gaussian method where the mean and covariance of chrominant color are calculated and the Mahalanobis distance to classified skin and non-skin threshold are measured. Then by using blob analysis technique, the centroid value is extracted and uses it as the position of the hand. Robotic platform can be controlled by a set of six instructions including stop, start, forward, backward, left and right based on the movement of the hand.