In Press                   Back to the articles list | Back to browse issues page

XML Print

Real-Time Detection of Common Objects and Humans with Relative Distance Measurement using pre-trained A.I Model for Visually Impaired Persons
Arpan Deyasi , Diptadip Barai , Pampa Debnath
Abstract:   (58 Views)
The present paper introduces a real-time intelligent assistive system that uses a pre-trained Artificial Intelligence (AI) model to estimate relative distance and detect common items/living beings and known people to help visually impaired people. A lightweight and accurate object identification framework based on YOLOv8 algorithm is used in the proposed system and tailored for embedded and portable devices for facilitating real-time detection. The system continually analyses live video data, recognizes objects and people within its range of vision, even moving towards the impaired with a significant relative velocity; and uses depth mapping or stereo vision algorithms to determine their relative distances. Relative distance is measured with accuracy varying from 98.66% to 99.8% for static objects and 98.67% to 99.47% for dynamic objects and both indoor and outdoor environments, with maximum variance of 1.21 and 5.76 respectively for static and moving objects, far superior compared to earlier results for more than 200 cm distance. With approximate 30% improvement in confidence score for static objects with 2.5% enhancement in mean average precision and 15% improvement in FPS, the proposed study presents a practical and scalable assistive solution that integrates AI perception and distance awareness to support independent navigation for visually impaired users.
 
 
Keywords: Pre-trained A.I Model, Relative Distance Measurement, Real-Time Detection, Visually Impaired People
Full-Text [PDF 1339 kb]   (8 Downloads)    
Type of Study: Research Paper | Subject: Integrated Circuits: Digital, Analog
Received: 2025/11/04 | Revised: 2026/08/12 | Accepted: 2026/06/29
Rights and permissions
Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Creative Commons License
© 2022 by the authors. Licensee IUST, Tehran, Iran. This is an open access journal distributed under the terms and conditions of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license.