Face Recognition for Human Following Robot using D435i Camera and Robot Operating System

Abstract

Most of the researchers only focus on people detection, which does not allow for recognizing specific individual. As a result, in environments with multiple people, this limitation makes it difficult for the human following robot to precisely track and follow a specific individual. Therefore, camera approach using Realsense D435i camera is implemented in this paper to recognize a specific human face as the first stage for the development of the human following robot. OpenCV is used as the underlying algorithm to recognize the human face based on the camera data obtained. In this study, a robot operating system (ROS) is utilized to integrate the camera data with the OpenCV modules for the face recognition process. The proposed model shows a remarkable performance, achieving an approximate accuracy of 70% in real-time recognition of the target individual by using camera. The experiment also has been evaluated and successfully recognize the target individual in the environment that have multiple persons. To enable the robot's data processing and task execution, the NVIDIA Jetson Nano, a cost-effective single-board computer with impressive processing capabilities, is employed and will be attached to the robot in the future works. In conclusion, the model can effectively be implemented on human following robot application using D435i camera.