Publication detail

Computer Vision Technique for Scene Captioning to provide Assistance to Visually Impaired

DAHIYA, D. ISSAC, A. DUTTA, M. ŘÍHA, K. KŘÍŽ, P.

Original Title

Computer Vision Technique for Scene Captioning to provide Assistance to Visually Impaired

English Title

Computer Vision Technique for Scene Captioning to provide Assistance to Visually Impaired

Type

conference paper

Language

en

Original Abstract

Commuting and other daily tasks are some of the biggest problems faced by visually impaired people. It becomes extremely difficult for visually impaired people to travel without any assistance and do the daily chores. The proposed method presents an automated and efficient way to tackle these problems by image processing based identification of familiar places like Restrooms, Pharmacies and Metro train stations. Symbols of these places are identified from an image by point feature matching using template detection method. A test image is compared with templates and accordingly output is obtained. The proposed method was tested on a varying database of images and achieved an accuracy of 91.67% for the proposed database. The experimental results present that this method for identifying Restrooms, Pharmacies and Metro train stations is efficient and very helpful for visually impaired people.

English abstract

Commuting and other daily tasks are some of the biggest problems faced by visually impaired people. It becomes extremely difficult for visually impaired people to travel without any assistance and do the daily chores. The proposed method presents an automated and efficient way to tackle these problems by image processing based identification of familiar places like Restrooms, Pharmacies and Metro train stations. Symbols of these places are identified from an image by point feature matching using template detection method. A test image is compared with templates and accordingly output is obtained. The proposed method was tested on a varying database of images and achieved an accuracy of 91.67% for the proposed database. The experimental results present that this method for identifying Restrooms, Pharmacies and Metro train stations is efficient and very helpful for visually impaired people.

Keywords

Estimate geometric transform; Image processing; Object detection; Point feature matching; Visually impaired assistive technology

Released

04.07.2018

Publisher

Faculty of Electrical Engineering and Communacation, Brno University of Technology

Location

Technická 3058/10, 616 00 Brno, Czech Republic

ISBN

978-1-5386-4695-3

Book

Proceedings of the 2018 41st International Conference on Telecommunications and Signal Processing

Pages from

372

Pages to

375

Pages count

4

URL

BibTex


@inproceedings{BUT148931,
  author="Dhruv {Dahiya} and Ashish {Issac} and Malay Kishore {Dutta} and Kamil {Říha} and Petr {Kříž}",
  title="Computer Vision Technique for Scene Captioning to provide Assistance to Visually Impaired",
  annote="Commuting and other daily tasks are some of the biggest problems faced by visually impaired people. It becomes extremely difficult for visually impaired people to travel without any assistance and do the daily chores. The proposed method presents an automated and efficient way to tackle these problems by image processing based identification of familiar places like Restrooms, Pharmacies and Metro train stations. Symbols of these places are identified from an image by point feature matching using template detection method. A test image is compared with templates and accordingly output is obtained. The proposed method was tested on a varying database of images and achieved an accuracy of 91.67% for the proposed database. The experimental results present that this method for identifying Restrooms, Pharmacies and Metro train stations is efficient and
very helpful for visually impaired people.",
  address="Faculty of Electrical Engineering and Communacation, Brno University of Technology",
  booktitle="Proceedings of the 2018 41st International Conference on Telecommunications and Signal Processing",
  chapter="148931",
  howpublished="online",
  institution="Faculty of Electrical Engineering and Communacation, Brno University of Technology",
  year="2018",
  month="july",
  pages="372--375",
  publisher="Faculty of Electrical Engineering and Communacation, Brno University of Technology",
  type="conference paper"
}