Iris tracking using extended object tracking

In the emerging field of hands free computer control the human eye has an extraordinary important role to play. In order to detect commands and actions communicated by the human eye, meaningful features are sought for. The iris provides high contrast, when compared to the sclera. Due to this fact the segmentation of the iris in the image of a face is fairly easy. This paper proposes an algorithm based on extended object tracking to trace the iris of the human eye. Based on a segmentation algorithm to find the boundary of the iris and together with simple model of iris motion an Unscented Kalman Filter (UKF) is used to track the iris. The measurements are incorporated using a greedy association model (GAM) to relate the measurement points to the extended object (the iris). We will show that the iris tracking algorithm is reliable and stable, even in presence of occlusions and reflections, e.g. when the person wears glasses.