Equipped with Monocular Depth Estimation and Intelligent Wake-Up Vision Based Tracking System for a Human-Following Mobile Robot

Abstract

With the advancement of technology and the development of human civilization, intelligent robots as “service robots” have been gradually integrated into the practical application of people's daily lives and the good assistants for people's family life, medical and health, entertainment, and social security. This paper designs and implements a novel vision-based intelligent human-following system for a mobile robot. To achieve human tracking using only RGB images, a monocular depth estimation network is combined with an object tracking network. Considering the speed, robustness, and accuracy in real applications, we optimized the network and added the target update strategy to achieve a good balance between speed and accuracy. Implemented on an NVIDIA Jetson AGX Xavier, the system achieved 97.8% accuracy on our customized dataset, with overall system performance averaging 5.57 FPS.