Loading...

Redesign and Reconstruction of a Smart Cane for Blind People to Detect Empty Seats and Social Navigation in Indoor Environments

Haghighat Joo, Mahdi | 2025

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 58189 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Taheri, Alireza
  7. Abstract:
  8. Visual impairments worldwide, especially among blind and visually impaired individuals, pose numerous challenges in mobility and personal independence. This research aims to design and develop a smart cane based on advanced computer vision technologies and deep learning algorithms, capable of real-time obstacle and environmental target detection, while guiding the user along safe and optimal paths that respect social etiquette. The proposed system utilizes an RGB-D OAK-D Lite camera, an IMU sensor, and a Raspberry Pi 5 microcontroller. Obstacle and object detection is implemented based on the advanced YOLOv8 architecture, and dynamic path planning is realized using the D* Lite algorithm. To identify collective human activities and adapt navigation to social contexts, the COMPOSER model—built on a multi-scale transformer architecture—was trained and integrated into the system. Tactile feedback for user notification is provided via three coin vibration motors. Field evaluations were conducted in both simulated and real environments with static obstacles and diverse human groups, demonstrating an overall system accuracy of 80%. The results indicate the system’s capability for safe guidance, accurate target recognition, and adherence to social norms in complex and dynamic scenarios. This research represents a significant step forward in developing intelligent mobility aids for visually impaired individuals, potentially improving their quality of life and independence
  9. Keywords:
  10. Path Planning ; Group Activity Recognition ; Obstacle Detection ; Smart Cane ; Social Etiquette ; RGB-D Camera

 Digital Object List

 Bookmark

No TOC