A Pedestrian Avoidance System for Visual Impaired People Based on Object Tracking Algorithm

Rui Shan, Wei Shi, Zhu Teng, Yoshihiro Okada

Abstract: Avoiding obstacles and pedestrians during the movement is a challenge for the visual impaired people. With the advancement of computer vision, some new technologies can be applied to improve this problem. In this paper, we proposed a new framework to realize such function based on the object tracking algorithm, FairMOT. Our framework detects the pedestrians around users and alarm users to avoid the collision between them. In our system, we use the 360-degree camera as the system input to record the users’ surrounding situations, and use the VR headset as the output device with a feedback application, that is realized using WebVR technologies. In this paper, we also discuss some experiments of the current system and some further improvement methods according to the results of experiments.