SWIR-LED-Based Road & Marine Condition Sensing for Automotive, Maritime and Intelligent Transport Systems
Executive Summary
Autonomous cars and unmanned surface vessels (USV) must “see” the first 0–2 mm of asphalt or water—exactly where ice films, oil sheens or condensed fog dictate friction and stopping distance. SWIR LEDs (1 000–1 900 nm) deliver three decisive advantages:
5–7× higher contrast between H₂O / ice and substrate than NIR-850 nm
20× less Mie scattering in fog & drizzle than VIS systems
Eye-safe, solid-state, AEC-Q102 & IEC 60945 qualified emitters already in volume production
This paper shows how Zhuhai Tianhui SWIR engines are being embedded today for lane-level friction mapping, port oil-spotting and polar ice-thickness estimation.
Physics & Differentiators
Band selection (tested)
1 310 nm – high water/ice absorption → ice vs water discrimination
1 550 nm – local atmospheric window, minimal solar background, good SNR in fog
1 650 nm – oil & diesel absorption shoulder, enables 0.1 µm film detection
1 390 nm (optional) – reference channel for common-mode drift cancellation
Polarisation add-on
Integrated wire-grid polariser (0.5 mm thin film) gives parallel & perpendicular channels; depolarisation ratio ρ = I⊥/I∥ directly correlates with surface micro-roughness → dry asphalt ρ≈0.12, ice ρ≈0.35, oil ρ≈0.45 (field data, –5 °C).
System Architecture
Module specs (AEC-Q102 Grade 0)
4-channel SWIR LED pixel (1 310/1 390/1 550/1 650 nm) in 3.5 × 3.5 mm LCC ceramic, 1 W total, 45 ° beam, rise-time <20 ns
Shared InGaAs linear array (320 × 1, 25 µm pitch) or single-point detector for cost-down variants
On-board NTC + TEC-less compensation algorithm (drift <1 count °C⁻¹)
MIPI CSI-2 / 100 BASE-T1 output, <250 mW average power @ 100 Hz frame
IP69K, –40 °C…105 °C, 1000 h 85 °C/85 % RH passed, 50 g vibration
Placement examples
Automotive: 2 modules flush-mounted in front bumper, 1.2 m baseline, 30° downward; fusion ECU receives SWIR, 77 GHz RADAR, 905 nm LiDAR and tyre-RF tags.
Marine: mast-head pod, 25 m range, synchronised with 24 GHz radar and AIS; second pod on bow for ice thickness.
Roadside: solar-powered edge unit every 500 m on bridges; LoRaWAN uplink to traffic control centre.
Field Results
Ice & snow (Heihe–Harbin highway, –18 °C, 0.3 mm black-ice)
SWIR contrast index (CI = (R_dry – R_ice)/R_dry) = 0.42 vs 0.07 at 850 nm
Fusion algorithm (SWIR + temp + humidity) reached 96 % F1-score; false-negative rate 1.1 % (vs 7 % for camera-only).
Oil spill (Port of Zhuhai, 0.2 µm diesel film)
1 650 nm channel showed 12 % drop in reflectance vs clean sea; detection limit 50 ppb (0.05 µm), alarm raised 3 min after leak began.
Fog penetration (South China Sea, visibility 350 m)
1 550 nm back-scatter 8× lower than 650 nm; USV kept 10 kn cruise when camera/LiDAR downgraded to 2 kn.
Added-Value Algorithms
Digital-twin friction map: each tyre pass uploads μ-peak slip value; Gaussian-process regression predicts μ 500 m ahead.
Ice-adhesion strength estimator: uses polarisation ratio + surface temp; validated against ASTM D3528 shear tests (R² = 0.81).
Oil-volume inversion: bi-directional reflectance model + wind-speed data → real-time spill volume within ±15 %.
Roadmap & Challenges
2025 6-channel chip (add 1 210 nm for tyre-rubber trace, 1 880 nm for snow density) + ASIC photon-counting ROIC
2026 Co-calibrated SWIR + FMCW LiDAR single aperture; goal: 5 W, 250 m range, size 50 × 30 × 25 mm
2027 Self-cleaning nano-coating (contact angle 110°) to cut maintenance 50 %; AI-powered predictive de-icing salt dosage (saves 20 % salt, CO₂ ↓12 kt yr⁻¹ on 1 000 km network)
Challenges: solar glint at dawn/dusk, salt crystal build-up, eye-safety certification for >1 W 1 550 nm versions—addressed by synchronous coded-pulse modulation and hydrophobic IP69K windows.
Zhuhai Tianhui SWIR Light-Source Update
In recent road-trial sections and port skid-deck tests, Tianhui’s automotive-grade 1 310 nm and 1 550 nm SWIR LEDs were used as active illuminators. Tier-1 suppliers report a ~15 % contrast gain for 0–0.3 mm black-ice and light-fog scenes, meeting minimum L3 autonomous-sensor recognition requirements. The modules have passed 1 000 h 85 °C/85 % RH ageing and are now in 5 000 km on-road accumulation testing; wavelength drift and thermal-management specs will be further optimised using open-road data.