Real-Time Face Tracking Under Illumination Variation

 

                                     Student: Chung-Min Chou    

                                     Advisor: Dr. Kai-Tai Song                                       

 ABSTRACT

                      Human-robot interaction is a key technology for a home service robot. In order to     interact with family members, a home service robot should track faces under illumination variation and have fast response to face motion. Color is an efficient cue for face tracking, but skin color is easily suffered by illumination and can make face tracking out of work. In this thesis, we propose an adaptive skin searching for tracking faces under illumination variation stably and rapidly. We use skin color detection and attentional cascade to detect frontal faces in real time. The face tracking and detection procedures have been implemented on an embedded image system developed by our lab. This real-time image system had integrated with a industrial PC and head motion control card to track faces. The