Motion Field Segmentation and Appearance Change Detection for Hand Tracking from the Ego Perspective

In this paper a novel method for moving object tracking is presented and applied in the context of hand tracking for mobile AugmentedReality (AR) applications. AR glasses-like devices come with an integrated camera capturing the field of view of the user. Hand gestures are the most intuitive interaction modality for manipulating AR contents and hand tracking is the first step towards robust gesture recognition. The presented method fuses motion segmentation and appearance change detection in a new way to track hands in front of complex backgrounds under varying lighting conditions without the need for previous color calibration. A comparison of this new algorithm with state-of-the-art tracking methods is conducted using a thorough evaluation methodology and challenging data sets containing different wiping hand gestures.