Operational Limits of Near-Infrared Face Recognition: The Critical Impact of Distance on Identification Accuracy

Raiz Karman, Indrabayu, I Putu Wahyu Kusuma

2025

Abstract

Robust face recognition in low-light conditions is crucial for security, yet the impact of varying stand-off distances on Near-Infrared (NIR) systems remains underexplored. This study quantifies this impact through a systematic evaluation of an NIR-to-NIR system. We constructed a dataset of 10 subjects at five controlled distances (4–20 meters) and implemented an end-to-end deep learning pipeline. Component validation showed that a Multi-Task Convolutional Neural Network (MTCNN) augmented with selective Contrast Limited Adaptive Histogram Equalization (CLAHE) achieved a superior 76.22% detection rate. The InceptionResNetV2 model, trained with Triplet Loss, achieved 80% overall identification accuracy and significantly outperformed a classical LBP+LDA baseline. However, performance was critically distancedependent, with accuracy dropping from 93% at close ranges to 55% at 20 meters. This degradation highlights the challenge of Low-Resolution Face Recognition (LRFR), limiting the system's practical range. Future research should target resolution enhancement via Face Super-Resolution (FSR) and the design of lightweight architectures for edge devices.

Paper Citation

in Harvard Style

Karman R., Indrabayu. and Kusuma I. (2025). Operational Limits of Near-Infrared Face Recognition: The Critical Impact of Distance on Identification Accuracy. In Proceedings of the 1st International Conference on Research and Innovations in Information and Engineering Technology - Volume 1: RITECH; ISBN 978-989-758-784-9, SciTePress, pages 53-60. DOI: 10.5220/0014269200004928

in Bibtex Style

@conference{ritech25,
author={Raiz Karman and Indrabayu and I Putu Wahyu Kusuma},
title={Operational Limits of Near-Infrared Face Recognition: The Critical Impact of Distance on Identification Accuracy},
booktitle={Proceedings of the 1st International Conference on Research and Innovations in Information and Engineering Technology - Volume 1: RITECH},
year={2025},
pages={53-60},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0014269200004928},
isbn={978-989-758-784-9},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Research and Innovations in Information and Engineering Technology - Volume 1: RITECH
TI - Operational Limits of Near-Infrared Face Recognition: The Critical Impact of Distance on Identification Accuracy
SN - 978-989-758-784-9
AU - Karman R.
AU - Indrabayu.
AU - Kusuma I.
PY - 2025
SP - 53
EP - 60
DO - 10.5220/0014269200004928
PB - SciTePress