Visible light positioning in underground coal mines using reference-point-adaptive GAN augmentation and SENet-CRNN
High-precision personnel positioning in underground coal mines is challenged by complex optical propagation and limited fingerprint samples. We propose an underground VLP method integrating generative adversarial network-based fingerprint augmentation and a SENet-CRNN positioning model. An underground visible-light channel model is established to characterize received power variations. The augmentation strategy is employed to enhance fingerprints. The SENet-CRNN captures spatial-channel-temporal features from received signal strength sequences. Experiments demonstrate an average positioning error of 0.0669 m and a 90th-percentile positioning error of 0.1196 m, exhibiting high positioning accuracy and feasibility for deployment in underground mine environments.
Authors
- lei jing (ORCID: https://orcid.org/0009-0006-4824-5332)
- Hongyang Zhang (ORCID: https://orcid.org/0009-0005-8972-3794)
- Lingsheng Kong
Institutions
- Tianjin University of Technology (CN)
Publication Details
- Journal
- Optical Engineering
- Published
- 2026-11-16
- DOI
- https://doi.org/10.1117/1.oe.65.11.118103
- Primary Topic
- Optical Wireless Communication Technologies
- Type
- article
- Field-Weighted Citation Impact
- 0.00