Evolutionary head pose measurement by improved stereo model matching

This paper presents a method to estimate the 3D pose of a human's head using two images input from two stereo cameras. The proposed method utilizes an evolutionary search technique of 1-Step genetic algorithm (1-Step GA) being improved to adapt for the real-time recogniton in dynamic images and a fitness evaluation based on a matching stereo model. Here, the head position and orientation are detected simultaneously using the brightness distribution and color information of the input images. Moreover, the facial features, eyebrows and eyes, are also recognized to increase the robustness of head pose measurement. An improved robust recognition for mobile robot has been proposed which is very effective to remove the influence of the camera motion.