Please use this identifier to cite or link to this item:
Title: Deep-learning approach for indoor image-based visible light positioning
Authors: Zhao, Guanliang
Keywords: Engineering::Electrical and electronic engineering::Wireless communication systems
Issue Date: 2022
Publisher: Nanyang Technological University
Source: Zhao, G. (2022). Deep-learning approach for indoor image-based visible light positioning. Master's thesis, Nanyang Technological University, Singapore.
Abstract: Precise location of personnel indoors is one of the crucial prerequisite of data analytics of workforce productivity as well as workplace safety and health. In this dissertation, we propose an enhanced indoor occupancy tracking system using optical camera communication (OCC) on top of conventional surveillance cameras. The proposed system is able to track those workers who carry unique infrared LED beacons indoor. OCC with infrared beacons is adopted for identity check, while the localization accuracy is enhanced using deep-learning-based human pose estimation to infer the footfall location. Particularly, a comprehensive footfall estimation algorithm empowered by deep-learning-based human pose estimation is presented, which provides precise footfall estimation considering real-time actions of person and occlusion patterns. In order to validate the accuracy improvement, the experiment is conducted in an open lab with area of about 30 m2 using single surveillance camera sensors. According to the experimental results, the localization accuracy is improved by 9.36% on average and 44.19% at far end, in comparison with conventional methods.
Schools: School of Electrical and Electronic Engineering 
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:EEE Theses

Files in This Item:
File Description SizeFormat 
  Restricted Access
19.31 MBAdobe PDFView/Open

Page view(s)

Updated on Sep 26, 2023


Updated on Sep 26, 2023

Google ScholarTM


Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.