Monocular vision-based collision avoidance system

Abstract

For the elderly people who have a low vision to safely navigate in unknown environments, the system should be developed to recognize where the obstacles in the scene are. In this paper, we present a vision system for obstacle detection, and implemented it on the Smartphone that provides real-time feedback to the user. In addition, various obstacles are localized using online background model, then viable paths to avoid them are determined by neural network-based classifier. Finally, the recognized results are verbally notified to the user through a visual interface. To demonstrate the effectiveness of the proposed method, it was tested on real indoors and outdoors with several environmental factors such as illumination type and complex structures. Then the results demonstrated the effectiveness of the proposed method.