Real-Time Multiple Faces Tracking with Moving Camera for Support Service Robot

Abstract

This paper proposes a real-time robot vision system to track multiple faces. This system supports service robots to communicate with consumers simultaneously. The Viola-Jones algorithm to detect faces early in the process, while the Kanade-Lucas-Tomasi algorithm is used to track detected facial features. The system follows the multiple human faces and synchronizes software and hardware to move the webcam in the middle position of the frame. Extraction of the center point from a set of faces as core information for controlling webcam movement. The challenge that can be overcome is that it can maintain multiple faces remains in the middle of the frame with various poses and the use of accessories. This system uses the PID controller which makes the webcam move in a fast to follow the face and maintain the stability and accuracy of actuator movements when speed increases. The experiments were done in the 42.052 frames per second as the maximum speed of system performance.