Integration of Optical Wireless Positioning and Deep Learning-Based Drone Detection

Alan Muñoz-Torres; Ismael Soto; Raul Zamorano-Illanes; David Zabala-Blanco; Cesar Azurdia; Zabih Ghassemlooy; Muhammad Ijaz

doi:10.1109/sacvlc67412.2025.11262018

Integration of Optical Wireless Positioning and Deep Learning-Based Drone Detection

Alan Muñoz-Torres, Ismael Soto, Raul Zamorano-Illanes, David Zabala-Blanco, Cesar Azurdia, Zabih Ghassemlooy, Muhammad Ijaz

School of Engineering Physics and Mathematics

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

This paper integrates optical wireless positioning (LED arrays + Cayley-Menger estimation with Bézier smoothing) and deep-learning drone detection. Atmospheric visibility significantly influences positioning accuracy, which improves from 23.2 cm MAE at 0.1 km to 11.8 cm at 20.0 km visibility for λ = 818 nm, where Bézier curves smooth trajectory errors while reducing path length versus Cayley-Menger results. YOLOv8 demonstrates slightly superior accuracy across most metrics, while YOLOv11 offers competitive performance with lower inference time, highlighting a trade-off between precision and computational efficiency. K-fold cross-validation enhances both models, with CV-YOLOv8 reaching 99.27% precision and 99.09% mAP50, demonstrating robust performance for indoor navigation.

Original language	English
Title of host publication	2025 South American Conference On Visible Light Communications (SACVLC)
Publisher	IEEE
Pages	1-6
Number of pages	6
ISBN (Electronic)	9798331556853
ISBN (Print)	9798331556860
DOIs	https://doi.org/10.1109/sacvlc67412.2025.11262018
Publication status	Published - 22 Oct 2025
Event	2025 South American Conference On Visible Light Communications (SACVLC) - La Paz, Bolivia, Plurinational State of Duration: 22 Oct 2025 → 24 Oct 2025

Conference

Conference	2025 South American Conference On Visible Light Communications (SACVLC)
Country/Territory	Bolivia, Plurinational State of
City	La Paz
Period	22/10/25 → 24/10/25

Keywords

Bezier curves
computer vision
deep learning
drone detection
indoor navigation
optical wireless positioning
trajectory planning
UAV localization
YOLO

Access to Document

10.1109/sacvlc67412.2025.11262018

Cite this

@inproceedings{897a4c3210964e3a82115fcc178b4238,

title = "Integration of Optical Wireless Positioning and Deep Learning-Based Drone Detection",

abstract = "This paper integrates optical wireless positioning (LED arrays + Cayley-Menger estimation with B{\'e}zier smoothing) and deep-learning drone detection. Atmospheric visibility significantly influences positioning accuracy, which improves from 23.2 cm MAE at 0.1 km to 11.8 cm at 20.0 km visibility for λ = 818 nm, where B{\'e}zier curves smooth trajectory errors while reducing path length versus Cayley-Menger results. YOLOv8 demonstrates slightly superior accuracy across most metrics, while YOLOv11 offers competitive performance with lower inference time, highlighting a trade-off between precision and computational efficiency. K-fold cross-validation enhances both models, with CV-YOLOv8 reaching 99.27\% precision and 99.09\% mAP50, demonstrating robust performance for indoor navigation.",

keywords = "Bezier curves, computer vision, deep learning, drone detection, indoor navigation, optical wireless positioning, trajectory planning, UAV localization, YOLO",

author = "Alan Mu{\~n}oz-Torres and Ismael Soto and Raul Zamorano-Illanes and David Zabala-Blanco and Cesar Azurdia and Zabih Ghassemlooy and Muhammad Ijaz",

year = "2025",

month = oct,

day = "22",

doi = "10.1109/sacvlc67412.2025.11262018",

language = "English",

isbn = "9798331556860",

pages = "1--6",

booktitle = "2025 South American Conference On Visible Light Communications (SACVLC)",

publisher = "IEEE",

address = "United States",

note = "2025 South American Conference On Visible Light Communications (SACVLC) ; Conference date: 22-10-2025 Through 24-10-2025",

}

Muñoz-Torres, A, Soto, I, Zamorano-Illanes, R, Zabala-Blanco, D, Azurdia, C, Ghassemlooy, Z & Ijaz, M 2025, Integration of Optical Wireless Positioning and Deep Learning-Based Drone Detection. in 2025 South American Conference On Visible Light Communications (SACVLC). IEEE, pp. 1-6, 2025 South American Conference On Visible Light Communications (SACVLC), La Paz, Bolivia, Plurinational State of, 22/10/25. https://doi.org/10.1109/sacvlc67412.2025.11262018

TY - GEN

T1 - Integration of Optical Wireless Positioning and Deep Learning-Based Drone Detection

AU - Muñoz-Torres, Alan

AU - Soto, Ismael

AU - Zamorano-Illanes, Raul

AU - Zabala-Blanco, David

AU - Azurdia, Cesar

AU - Ghassemlooy, Zabih

AU - Ijaz, Muhammad

PY - 2025/10/22

Y1 - 2025/10/22

N2 - This paper integrates optical wireless positioning (LED arrays + Cayley-Menger estimation with Bézier smoothing) and deep-learning drone detection. Atmospheric visibility significantly influences positioning accuracy, which improves from 23.2 cm MAE at 0.1 km to 11.8 cm at 20.0 km visibility for λ = 818 nm, where Bézier curves smooth trajectory errors while reducing path length versus Cayley-Menger results. YOLOv8 demonstrates slightly superior accuracy across most metrics, while YOLOv11 offers competitive performance with lower inference time, highlighting a trade-off between precision and computational efficiency. K-fold cross-validation enhances both models, with CV-YOLOv8 reaching 99.27% precision and 99.09% mAP50, demonstrating robust performance for indoor navigation.

AB - This paper integrates optical wireless positioning (LED arrays + Cayley-Menger estimation with Bézier smoothing) and deep-learning drone detection. Atmospheric visibility significantly influences positioning accuracy, which improves from 23.2 cm MAE at 0.1 km to 11.8 cm at 20.0 km visibility for λ = 818 nm, where Bézier curves smooth trajectory errors while reducing path length versus Cayley-Menger results. YOLOv8 demonstrates slightly superior accuracy across most metrics, while YOLOv11 offers competitive performance with lower inference time, highlighting a trade-off between precision and computational efficiency. K-fold cross-validation enhances both models, with CV-YOLOv8 reaching 99.27% precision and 99.09% mAP50, demonstrating robust performance for indoor navigation.

KW - Bezier curves

KW - computer vision

KW - deep learning

KW - drone detection

KW - indoor navigation

KW - optical wireless positioning

KW - trajectory planning

KW - UAV localization

KW - YOLO

U2 - 10.1109/sacvlc67412.2025.11262018

DO - 10.1109/sacvlc67412.2025.11262018

M3 - Conference contribution

SN - 9798331556860

SP - 1

EP - 6

BT - 2025 South American Conference On Visible Light Communications (SACVLC)

PB - IEEE

T2 - 2025 South American Conference On Visible Light Communications (SACVLC)

Y2 - 22 October 2025 through 24 October 2025

ER -