Implementation of a Blind navigation method in outdoors/indoors areas

TL;DR

This paper presents a monocular camera-based navigation system for visually impaired individuals, combining visual SLAM and object detection to guide users safely in indoor and outdoor environments with high accuracy and real-time obstacle alerts.

Contribution

It introduces a novel integration of ORB-SLAM and YOLO algorithms for blind navigation, enabling real-time route guidance and obstacle detection with precise positioning.

Findings

System runs at 30 fps on a laptop camera.

Guides users within 11-meter routes with 8cm accuracy.

Alerts users of deviations over 60cm from the route.

Abstract

According to WHO statistics, the number of visually impaired people is increasing annually. One of the most critical necessities for visually impaired people is the ability to navigate safely. This paper proposes a navigation system based on the visual slam and Yolo algorithm using monocular cameras. The proposed system consists of three steps: obstacle distance estimation, path deviation detection, and next-step prediction. Using the ORB-SLAM algorithm, the proposed method creates a map from a predefined route and guides the users to stay on the route while notifying them if they deviate from it. Additionally, the system utilizes the YOLO algorithm to detect obstacles along the route and alert the user. The experimental results, obtained by using a laptop camera, show that the proposed system can run in 30 frame per second while guiding the user within predefined routes of 11 meters in indoors and outdoors. The accuracy of the positioning system is 8cm, and the system notifies the users if they deviate from the predefined route by more than 60 cm.