EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Artificial Intelligence of Things-Enabled Smart Assistive Framework using Explainable Vision Transformer for Visually Challenged People.

Ehab Bahaudien Ashary

Computer and Decision Making: An International Journal · 03.11.2026 · 3 (6S) · 3285-3298

DOI: 10.1922/comdem.v3i6s.19067 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Computer and Decision Making: An International Journal, online erschienen am 03.11.2026. Verfasst von Ehab Bahaudien Ashary. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Artificial Intelligence of Things (AIoT) integrates Artificial Intelligence (AI) with the Internet of Things (IoT) to design intelligent and automated systems. It finds useful to derive useful assistive technologies for visually challenged individuals via the incorporation of smart sensors, AI, and interconnected devices. Visually impaired persons (VIPs) face many complications in their daily lives, and technological mediators can aid them in meeting their challenges. In addition to other useful technologies, object detection is a computer technology correlated with computer vision (CV) and image processing that recognizes and defines objects either in indoor or outdoor environments. VIPs can use the object detection technique to identify hindrances and identify services to offer protected and informative navigation. Numerous technologies have been presented to help VIPs in their indoor navigation. In recent times, deep learning (DL) based methods have been trained with many object images that are closely linked with persons with disabilities. DL methods were the most effective technique for image processing till now, which are according to an artificial neural network (ANN) with deep structure. In this study, we develop an Intelligent Object Detection System Using a Vision Transformer and Deep Representation Learning (IODS-VTDRL) methodology. The key purpose of the IODS-VTDRL method is to derive an intelligent object detection approach to help VIPs with disabilities navigate their environment safely and independently. At first, the image pre-processing step involves numerous levels like image acquisition, noise removal, image resizing and normalization, and contrast enhancement to eliminate noise and make image input data more evident. Followed by, the YOLOV12 approach is applied for object detection to locate, classify, and identify objects within an image. For feature extraction, the presented IODS-VTDRL model executes the vision transformer (ViT) technique. At last, the deep belief network (DBN) method is executed for the recognition model. To improve interpretability, the proposed model applies Grad-CAM which produces visual explanations for decisions. The effectiveness of the IODS-VTDRL system is certified by the complete analysis utilizing the benchmark image database. The mathematical outcome displays that the IODS-VTDRL methodology has greater scalability and performance under several evaluations over the recent approaches.

Zitieren

APA

Ashary, E. B. (2026). Artificial Intelligence of Things-Enabled Smart Assistive Framework using Explainable Vision Transformer for Visually Challenged People. Computer and Decision Making: An International Journal, 3(6S), 3285-3298. https://doi.org/10.1922/comdem.v3i6s.19067

Vancouver

Ashary EB. Artificial Intelligence of Things-Enabled Smart Assistive Framework using Explainable Vision Transformer for Visually Challenged People. Computer and Decision Making: An International Journal. 2026;3(6S):3285-3298. doi:10.1922/comdem.v3i6s.19067

BibTeX
@article{Ashary2026,
  author = {Ashary, Ehab Bahaudien},
  title = {Artificial Intelligence of Things-Enabled Smart Assistive Framework using Explainable Vision Transformer for Visually Challenged People},
  journal = {Computer and Decision Making: An International Journal},
  year = {2026},
  volume = {3},
  number = {6S},
  pages = {3285--3298},
  doi = {10.1922/comdem.v3i6s.19067}
}
Kostenlos starten →