Design and implementation of an Embodied face recognition and tracking algorithm

This paper mainly designs an algorithm for face recognition and tracking and implements the algorithm using ROS robots. Given the high hardware requirements of the actual ROS robot in the target tracking process, the accuracy is not accurate, and the recognition and tracking speed is slow. This article tries to use a combination of Haar-like, key point tracking, and optical flow tracking to improve the above problems. Secondly, this article uses camshaft color tracking to address poor image transmission quality under the ROS framework, the large difference in light difference, inaccurate depth information, and other problems caused by the loss of recognition and tracking targets solve. Finally, the algorithm is implemented on the robot, and the accuracy and advancement of the embodied face recognition and tracking algorithm in this article are verified through experiments.