Citation

Rahman, Tanzeel Ur and Li, Guijun and Khan, Adnan Daud and Ali, Farman and Iqbal, Saeed and Nuaim, Ahmed Al and Ouyang, Zhengbiao and Alruwaili, Omar and Kamal, Shahid (2026) Dual-beam differential optical imaging for atmospheric turbulence mitigation in free-space propagation. Optics Express, 34 (10). p. 17770. ISSN 1094-4087

Text oe-34-10-17770.pdf - Published Version Restricted to Repository staff only Download (8MB)

Abstract

Atmospheric turbulence imposes a fundamental barrier to reliable object detection in consumer-grade free-space optical (FSO) systems, degrading performance in applications such as drone navigation, smart surveillance, and portable satellite terminals. Current approaches, exemplified by the hybrid adversarial contrastive framework (HACF), operate in the post-acquisition domain, treating turbulence as irrecoverable stochastic noise and relying on statistical pseudo-label refinement to mitigate its effects. We introduce differential beam imaging (DBI), a physics-driven, acquisition-layer paradigm that transcends this limitation by physically isolating object reflectance at the moment of signal capture. DBI exploits the spatial correlation of atmospheric turbulence through simultaneous dual-beam probing; one beam illuminates the object, while an adjacent reference beam captures only the turbulence field. By differencing these measurements, DBI cancels turbulence at its source, preserving the object’s structural integrity before any information loss occurs. We formalize and solve four non-incremental scientific problems: (P1) establishing the wave-optics-based spatial correlation regime for valid cancellation, (P2) optimizing beam separation under turbulence crosstalk trade-offs, (P3) designing a Siamese-cross attention network (SCAN) that learns turbulence-invariant features from differential image pairs, and (P4) co-designing a real-time, low-power (<1 W, <100 cm3) hardware-software stack for consumer deployment. Simulation-based validation demonstrates that DBI achieves a Turbulence Cancellation Efficiency (TCE) exceeding 2.3 and maintains a mean average precision (mAP) of 64.1 under moderate turbulence ( m ), outperforming HACF by 35% in detection accuracy while operating at 30 FPS. This work establishes a new class of differential optical systems, shifting the paradigm from algorithmic robustness to physical-layer intelligence for robust perception in dynamic optical environments.

Item Type:	Article
Subjects:	Q Science > QC Physics > QC350-467 Optics. Light
Divisions:	Faculty of Computing and Informatics (FCI)
Depositing User:	Ms Suzilawati Abu Samah
Date Deposited:	05 Jun 2026 04:08
Last Modified:	05 Jun 2026 04:08
URII:	http://shdl.mmu.edu.my/id/eprint/16006

Downloads

Edit (login required)