Real-Time Localization and Tracking System with Multiple-Angle Views for Human Robot Interaction

摘要

In this paper, we introduce an approach for robot head mechanism using a parallel mini-manipulator model to establish a highly efficient multiple angle view functionality for robot vision systems. In order to demonstrate the abilities of the structure, we implement the well-known Eigenfaces and Haarcascade algorithms for real-time tracking, localizing, and recognizing problems on human-robot interaction (e.g grasping and manipulation in autonomous robots). The advantages of these algorithms are the guaranty of high-speed face detection with also high accuracy, which proposed to be robust and efficient. The system could be integrated into Robot Operating System (ROS) so that we can control many kinds of peripherals signal more effectively. Furthermore, ROS provides our system with scalability for further development more effectively.