Development of walking support device for visually impaired people based on client-server system

Abstract

When a visually impaired person goes out, all kinds of accidents such as falls and contact with other people are expected. Therefore, using object recognition AI, we constructed a system that detects obstacles around pedestrians, obstacles around the pedestrian, etc. that pose a danger to pedestrians in real time from the camera image and alerts the user to that There is. However, in this system, all the processing was performed by a single microcomputer, so it is difficult to expand the functions due to problems such as delay and lack of memory. Therefore, we propose a client-server system that uses two microcomputers for the client and server, and outsources the object recognition, which requires the most processing, to the server side, thus enabling the addition of extended functions such as deviation detection. As a similar paper, “A new approach to wearable image recognition system to assist visually impaired persons “(Shiwei Chen, Dayue Yao, Huiliang Cao * and Chong Shen *MDPI, Published: 15 August 2019)[1]. It is a system that sends the information from the sensor to the cloud server and returns the processing result performed by the server to the client, but this system can be used by connecting to the net. Since the system we made uses ROS, it has the disadvantage that the system can only be used within the same LAN. Therefore, we propose a client server system using VPN connection.